Notes on the Roman Forts at Rough Castle and Wester- Wood, Wit Hpostscripta Georgb Si Y B E

II. NOTES ON THE ROMAN FORTS AT ROUGH CASTLE AND WESTER- WOOD, WIT HPOSTSCRIPTA GEORGB SI Y B E. MACDONALD, K.C.B., LL.D., D.LiTT., P.B.A., P.S.A.ScoT. The following communication may most fitly be regarded as a continuatio f thao n t e Societmadth o t e a yyea o regardin ag r e th g forts at Old Kilpatrick and Croy Hill. The motive underlying the investigatio e methods similarth wa nd an ,s employed were identical. Thasayo t eacn s i ,i t h cas ea stud f previouyo s account d revealeha s d a poin r pointo t f speciao s l difficulty whic ha littl e spade-work might reasonably be expected to elucidate, and in both the principle of aiming at a limited objective was r-igidly adhered to. Here and there clues which unexpectedly presented themselves were followed up. But in the process d indeean , d throughout, great care was take o avoit ne th d unnecessary disturbance of anything that might prove valuable as evidence in the event of a really searching exploration being carried e mor th r oules o en i ts distant future. Permission to cut a few trenches at Rough Castle was most courteousl . DouglaT r yM sgrantey b Wallace e m d ,e th facto n o r Callendar estate, acting on behalf of Mr Forbes, the proprietor, while Misd Mran s Drysdale were equally kin consideratd dan t Westerwoodea . e actuath r l Fo diggin s fortunatelwa gI y abl o securt e e serviceth e s r AlexandeM of d ralread ha Mann o y wh ,o effectivelhelpes e m d y elsewhere. Circumstances unfortunately compelled me to draw largely on the kindness of friends for assistance in the work of supervision. . CurieO . MA r, C.V.O., paid repeated visit boto t s Johr h M sites nd an , Mathieson spent three days at Westerwood, where some excellent photographs were taken for me by Mr Curie and Mr J. S. Richardson. abls folloo t ewa I w y thin a tim I what s wa goina e wa n whet o g n I was laid aside by illness. Once again, however, my chief obligation

Samuer M o it s l Smith grudgeo wh , d neither tim r troubleno obtaio et n Early1 Christian Monuments of Scotland, part iii. p. 418, fig. 435. 2 Ibid., part iii . 428p . , fig. 448. 244 PROCEEDING E SOCIETYTH F O S , MARC , 1933H13 . informatioe th f whicho n virtuan i I s wa. searchn i wa sl e chargH . e e operationoth f s during some critical stage t Bouga s h Castld an e during their whole cours t Westerwooda e r Calder'M . e s th pla f o n latter fort, which appears as fig. 13, is a reproduction of one drawn by Mr Smit n measurementsh ow fro s mhi , verifiee morth t ea d difficult points by Mr Mathieson. Figs. 6, 7, and 8 are also based on Mr Smith's measurements and sketches. The cost of the enquiries was covered by fro 4 a grane Society'£2 mth f o t s Excavation Fund.

. ROUGI H CASTLE.' To those whose memory carries them back to 1903 a visit to the Rough Castle of to-day is a somewhat depressing experience. The excavation f thao s t year aroused much public interest while they were in progress, and an appeal that they should be ' left open' was acceded to with a literality that has had unforeseen and disconcerting results. No steps whatever were taken to replace earth that had been moved or to protect the not inconsiderable remains of masonry that were exposed. The upcast, allowed to lie where it had been thrown, is now covered wit hgrowta h that render t indistinguishabli s e fro wore mth k Romanse oth f , whil fragmente eth buildingsf so , whe t disintegratenno d by the weather, have been maliciously demolished by human hands, and the stones utilized for the building of hearths by vagrants who have sheltered within the ramparts. So far, therefore, as the interior is concerned e tas th f verifyin, o k r amplifyino g g earlier results ha s been rendered extraordinarily difficult. The Bath-house (No. 4 in figs. 4 and 11), which was in very fair condition when first uncovered, present sspecialla y melancholy spectacle. There is nothing to recall the photographs of 1903, and even the plan

is virtually useles guida s s a excellenplan e a upo th s n i t spote nth Ye .t 2 excavatore oneTh . se tru seeth ideo e n f mo hav o at characted ha e r e ruineoth f d buildin whicn go h thed lighted t i sayyha d s an ,muc h for the conscientious accuracy of Mr Mungo Buchanan's work that, despite this heavy handicap, he should have produced a drawing which, taken as a whole, fits quite admirably into the setting of present-day knowledge. As no attempt to decipher its meaning is made in the original Report, a digression on its interpretation may not be irrelevant here. As a matter of fact, the lessons learned at Mumrills make it extremely eas s reproducereado i yt t I . e onl th fign , i dy1 . changf o e originae 1Th l Repor Rougn o t h Castle, whic quotes hi d passim whan i t follows, wil foune b l d in Proceedings, vol. xxxix. (1904-5) . 442pp , ft. Detailed reference givee sar onln i yvera w yfe exceptional cases. ' Proceedings, vol infra,d . ft. xxxix6 an . , 3 48 d d figs.an an (1904-5)2 4 . 48 . pp , ROMAN FORTS AT ROUGH CASTLE AND WESTERWOOD. 245

10 FEEO 3 T O 2 IO

Fig. 1. Plan of the Bath-house. 246 PROCEEDING E SOCIETYTH F O S , MARC , 1933H13 . substance bein e omissiogth dottea f no d line whic expressls hi y stated to be "conjectural." The descriptive term 'Retaining Wall,' which refer thio t s s lines alsha ,o been lefe otheth t outn r O handw . fe a , letters have been added to facilitate reference. As in the case of so many other castella at home and abroad, the entranc bathe fro s th Militare mso th wa et y Way, whic pasn hra t them on the north. A gravelled courtyard gave access on the east to a paved apartment (A), which served as a combined Apodyterium and Frigidarium. Mr Buchanan's "clay bed" doubtless represents the partition wall. Although 110 trace of a cold bath or of a basin for douching has survived, the former presence of one or, possibly, of both is adequatel 'gutter,e y th vouche y b e seer 'b whic dfo no t outsid s hi e whicd an onln e , walhb ca th y a lhav e been intendee o carrth dt f yof waste-water from the Frigidarium. B and C, both hypocausted rooms, were respectively the Tepidarium and the Caldarium. D, from which the furnace juts out into C, was the Prsef urnium; and E, which projects on the west and is likewise hypocausted, was the Sudatorium. The drain issuing from beneath the furnace calls for special notice. Indeed, a desir s examino t ewa t i e thi e sfirs th tha o tt n i instancte m d le e retur Rougo nt h Castle. Thaanalogous i t i te puzzlin th o t s g conduits which were observed at Mumrills is proved by the course it follows, coupled with the account given of it in the Report, where it is said to have been " 5 inches wide and 9 inches deep, covered with flagstones." The circumstance that " when opened it was found to be entirely choked with hard soot" may be thought to lend some countenance to the idea of a discarded ventilating-apparatus, for choking with soot is not quite the fate that one would anticipate for a channel that conveyed water. e otheOth n r hand s certain—i t i , s abl verifo wa t eI e poinX th y t a t last October—that it was linked up with the ' gutter' which drained the Frigidarium. This is new and important evidence, which will have reckonee b o t dfutury witan n ehi endeavou determino rt purpose eth e of such conduits. Nor is it only in regard to the allocation of the rooms that the plan is instructive. It also throws light on the character of the hypocausts, confirming in a striking fashion the inferences drawn at Mumrills. B and C were heated by pillared hypocausts. In the immediate neigh- bourhoo e furnaceth f o d , wher e firth ee woul fierceste b d e pillarth , s were of brick, although, as the photographic record (fig. 2) shows, one o t f soma the d em ha tim r otheo e r been replace a singl y db e stone. Elsewhere ,l eventaal t s durin finae gth l phase floore th , s overhead dha been supporte y blockb d f roughlo s y built masonry. Neverthelest i s woul mistaka e db calo et l these hypocausts channelled. Accordino gt ROMAN FORT T ROUGA S H CASTL WESTERWOODD AN E 7 24 . the Report, "the pillars placed next to the walls stand clear of them by inches3 r o 2e spac th , e between being completely clogged with soot, among which were many broken pieces of tile." This means that the hot carries e wallinside th th wa f sp r o e du abovai e throug h'jacketinga ' of box-tiles, warming the rooms by radiation. It is said by the excavators tha Sudatoriue th t "similas wa ) arrangement.mn i (E r t fig1 .Bu " shows that a differencether s wa e tha d e hypocausan th t, t thers ewa

Fig. 2. South end of Caldarium, showing mouth of furnace. channellea d hypocaus strice th n t i termt e supportine sensth Th f .eo g blocks of masonry are larger, and where they approach the walls they actually abut upon them r coulai .t d Thuho ris e eende sth onl th f st o y a the narrow channels, and therefore in streams •which would be quite ineffectiv r radiatinfo e g purposes. t I mus t have been introduced directly inte room furnace oth e Sudatoriu th Th . f o e s evidentlmwa y small d disappeared, ha sinc l t signi al e f o s . This point charcoao t s l fuelling.1 It is well known that Bath-houses were peculiarly liable to deteriora- 1 For a fuller discussion of this and other general points referred to here, see Proceedings, vol. Ixiii (Mumrills). . ff (1928-29) 7 44 . Arch.d pp ,an , , vol AeL,S. .h viii . 214t 9pp , ff. (Chesters). 248 PROCEEDINGS OF THE SOCIETY, MARCH 13, 1933. tion. We are thus prepared to learn from the Report that "evidence of alteration additiond san s appear throughou entire th t e building.t Bu " we can go a good deal further, if we bear in mind that in fig. 1 "the portions shaded black indicate what exists of the earlier walling," and that these had been reduced to the basement course or even to the foundations. Three main periods see o reveamt l themselves clearly. The fragmentary , takewalb a ln together wit e foundatiohth n that projects eastward l probabilital s n i int , oB y define e originath s l area centrae oth f l portio establishmente th f no , whic largests thes it t hwa na .

Pig. S. Reconstruction in the Caldarium.

It is impossible to say what change, if any, was made in E, when rebuild- toog in k e openinplac th e seconhow, t th D ea f d -go d an period, C , B . ever, were reducee B cas th f eo n i n widtdi t bu h, fro15 fee3 m2 o t t compensatio s provide e additioe nortnwa th th fee7 y n hb o f do t tno its length at the expense of the Apodyterium. During the third period dimensione th s see havmo t e remained unaltered plane Th ., therefore, gives no help there, and we have to turn to one of the photographs of 1903 for proof of complete reconstruction in a singularly slipshod style (fig. 3) . To make the other points to be dealt with intelligible, the plan of the fort (fig. 4) is here reproduced in the exact form in which it appeare e originath n e puzzlindi th l f Reporto e g featureOn . s which a study of it brings out is the occurrence of "mounds" in curiously unexpected places. In -walking over the ground it is impossible to PLAN, or THE ROMAN TORT ON THE ANTONINC VALLUM.

AT ROUGHCASTLC, STIRLINGSHIRE. 1904. !!K^^

WMMt^itti^

L;ua'iJ'J''Ujiu

PRO r MTHo itE SURFACE., THROUG MILITARE TH H Y WAY. SCALE. Buehtuttm, D»t.

Fig. 4. 250 PROCEEDINGS OF THE SOCIETY, MARCH 13, 1933. avoi e suspicioth d n thae excavatorth t s hav o ofteto e n been misled by appearances. The word, as they use it, inevitably conveys the impression of a mass of earth that has been deliberately heaped up to serve some defensive purpose. But it is only to two of the " mounds " marked on the plan (or mentioned in the text) that this definition applicable.y wa y seeman n 1i s Thus e "largth , e moun r platformo d , 100 feet long by 50 feet broad," which is shown "projecting outward e sout th e easo f nth ho t gateway whicd "an describes hi "s da special defence,l probabilital n i s i " y nothinupcase th t tgbu fro e ditchesmth , spread over a wide area with the express intention of preventing the accumulation from reachin heighga t that might have interfered with the outlook of the defenders upon the ramparts. Its top is on a level with the margin of the outer ditch, and it is only in relation e slopinth o t g ground beyon fairln t ca i thade spoke b ys t i a t f o n a moun t alla d . This conclusion, arrive t froa d ma consideratio f no what is to be seen upon the surface, was confirmed by the observations made at the one or two points at which we opened it up last autumn. The eart beed ha hn t i lookethrow f i s da n down quite loosely. There was no indication of the packing or ' ramming' which one would naturally associate wit theore hth "a specia f yo l defence." Moreover, that theory is hardly consistent with wha sais i t anothen di r passag e Reporth f eo t a roa o t d s whica h comes fro e eastmth , rises over "the traverse" and " is carried across the platform."2 quesn i s f thio t wa s t roaI d thamounde openee th w e t p Th u d. t quit e no searcresul th s e f conclusivewa ho t , perhaps because th e ground had previously been disturbed. But there was sufficient pre- sumptive evidence to satisfy both Mr Curie and myself: over a width of 18 feet the soil was markedly different from that on either side, fee8 1 s exactli t d e an widtyth he Militar usuath r fo l y Way f whic,o h this road was a branch. The branch had first attracted my attention more than twenty years ago. 1 pointed out then that it was later tha e Militarth n itselfy yWa , from whic t hi parte d compane th t ya eastern entranc e annexeth o e clearest e th , t proof being furnished face byth t (ascertaine e originath y b d l excavators) tha southere th t n

These are the one in front of the short ditch at the foot of the slope on the north-west, and ' traverse e th 1 ' outsid norte th f heo gate .surl al feeNoI et o a lrtha d lattee t th tf theso rno s ei largely natural. more Th 2 e easterl e " stonth s i y e t pari pave f o t d way" whic shows i h fig n , skirtinno 4 . g the outside of the annexe ditch. It must have been very visible two hundred years ago, for comparison with Gordon's plan shows that it is to this road that he is referring when he says : Circumstance "On vers i e y remarkabl t thiea s Fort; namely, thasame th t e Free-stone Wall already mentioned, seems by its Foundation here, to have surrounded the whole Castellum" (Itin. Sept., p. 59). He obviously mistook the bottoming of the road for the stone base of the "Vallum, whic lowese believee hh th e b t o coursdt freestona f eo e wall. ROMAN FORTS AT ROUGH CASTLE AND WESTERWOOD. 251 entrance must be an afterthought as the ditch in front of it had been filled in to provide a passage for the new approach.1 The object was, e inconvenientlth e f whab o turned o ot d fha t ri t courset dou yge o t , steep gradient betwee Rowae e westh nth fore td ngatth an tTre f eo e Burn, whic crosses e hardl th hforda wa f y wels o dyb a , s lesa l s trying climb on the farther side of the stream. On the alternative route, which was in every way much easier, a bridge took the place of the ford. All this remains broadly true. But an important modification is called for. Sinc I ewrot 1911n ei I hav, e collected evidence from other forts—it would hardly be appropriate to set it out here—which makes t virtualli y certaisingla e Militars a th ided eny f ol ao tha e yWa th t line, passing through each of the ' stations' (except Bar Hill), must be abandoned, and that there was a series of 'loops,' so arranged that s possiblwa t i o travet e l froe Clyde Fortth mth o et h withouy an t break in the journey. That it should have been so is just what might have been expected from Roman common sense. Many, if not most, conspicuoun o forte oe th f ar s s heights t woulI . d have bee nserioua s inconvenience f i traffi, c intende r somfo d e particula f theo e mon r could only have reache s destinatioit d n afte successioa r f toilsomno e d unnecessaran y ascents t leasa n a certaiI t . n numbe f instanceo r s —and Rough Castle ia cass n point—thi e e detou s nowa tr merely easie t shorterbu r . Nevertheles t seemi s s bette o spea loopst ' r f o k ' reservo t d e teran th em Militar' y Way ' e roafoth rd that woule b d used by detachments patrolling the frontier. The extent to which the interpretation of 1911 requires to be amended will be apparent from a glance at fig. 5. The road which is there seen approaching the south gate of the fort from the south-east, and which (as we learn from the Report) the excavators traced backwards in that direction for some 200 feet,2 was the end of a ' loop' which had left the main Way some distance farther east and had skirted the souther nswame edgth f eo p tha behiny l tla likelihoo Wallal e dth n I . d t "wai carro t s e 'loopyth ' tha e bridg th toriginalls ewa y thrown over the Burn. Accordingly, when it was decided to seek relief from the trying conditions that the use of the via principalis had been found to involve (A), there was no need to continue the new branch (B) beyond its natural point of junction with the already existing 'loop.' The two merged into one another immediately outside of the south e forth gat tf t o e(figwilI e remembere . b l .5) d thae excavatorth t s

1 The Roman Wall in Scotland (ed. 1911), p. 229. e Reporth 465. n I (p t ) hundreo the "y tw ysa d yards." That, however impossibles i , o Tw . hundre2 d yards would have carrie railwae dth theo t yn m o linint d brickworke ean oth e th n so farther side. 252 PROCEEDINGS OF THE SOCIETY, MARCH 13, 1933. note extensivn da e cobbled area just here. Thi precisels si y what might have bee nspoa looket a t r wherdfo e stream f traffio s cy metma t I . be added that s figa , indicates.5 e are e sout th ,fronn th ai f ho t gate e annexoth f beed eha n simila.rly treated.1

Our attention was next turned to the annexe itself. I was anxious to test the validity of the hypothesis which I had put forward in 191 accouno 1t e appearanc th r fo t f threeo e ditche s easterit n o s n side e excavatorTh (fig. 4) d suppose. ha s d tha l thre al td bee g ha edu n simultaneously, and that the two enclosed spaces beyond the innermost were "raised platforms.," to be manned and held by detachments

earliee FigTh . r .5 late (Ad Militar ) t Roug)an a r (B y hyWa Castle. of the garrison, who would thus be posted outside of the ramparts altogether. Findin t i impossiblg o reconcilt e e this scheme with any intelligible system of defence, 1 suggested that the three ditches, instead of being contemporaneous, really represented three consecutive stages in the history of the enclosure, the second and third stages being each marke appreciabln a y db e reductio s sizeit n .ni Tha e constructioth t n of the branch road had coincided with the opening of the second stage —that is, with the first reduction—was indicated by the ruthless manner whicn i e diggere seconth hth f o sd ditc d dealha h t wit e Militarhth y Way. They had cut right through it, as if it were henceforward to be of no value as a thoroughfare.2 One difficulty confrontin e vieth gw thas jusha tt been summarized was that, on Mr Buchanan's plan, the outermost ditch is shown stopping 1 The limits of these two cobbled areas, as indicated in fig. 5, are naturally very tentative. To thep mma accurately would have involve strippine dth mucf go h turf. ' The Roman Watt in Scotland (ed. 1911), pp. 229 ff. ROMAN FORTS AT ROUGH CASTLE AND WESTERWOOD. 253 abruptly on the south side of the Military Way without re-appearing alwayd ha beyon t I s seeme . possible dit m o dt e that this e mighdu e b t to a misapprehension on the part of the excavators. The work of last autumn proved that the suspicion was well founded. Several cross- sections wer t betweeecu e norte th nth d he Militar an sidth f y eo yWa Antonine Vallume firs f th theseo t n I ., whicyard w s onlfe wa hsya from the edge of the road, the ditch duly showed itself in its normal dimensions littlA . e farthe , howeveron r t begai , groo nt w narrower shallowerd an s proportionit , s diminishing rapidly until, afte suddea r n bottom e incheleve0 e th 1 ris th f f o eln o s i t becam i , e little more thana mere dent upon the surface.1 By the time the kerb of the Antonine Vallum was reached disappeared ha t ,i d entirely. After the continuance of the ditch had thus been established, it seemed worth while making some search for tracerampare th f so t which had presumably run behind it. These turned out to be somewhat indefinite. Nevertheless they seemed to me unmistakable. In each of three sections, cut across the probable line, there was found a bottoming of stones, loosely laiaboud dan fee6 t t hardln broadca t I y. have been designe o support d y structuran t e more elaborate tha n earthea n n takee b r grantey moundnfo ma t dI . tha e upcasth t t fro e ditcmth h furnishe necessare dth y material rampare Th . t behin centrae dth l ditch seeme similae b o d t structure same n ri th t e me .d ha Certainl o tw e yth fate. As they lost their value with the successive contractions of the e annexeareth f ao , they were roughly levellemateriae th d dan l spread over the ground that had been abandoned. That is why, in the words of the excavators, both of " the platforms on the east defence . . . show a decided increas heighn ei t where facin south-east.e gth " Ther littls ei e addee b o t d regardin e thir gth r innermos o d t rampart e onle th , yon marke e stons suca dTh n fig o h e. .4 bottomin s perhapwa g s more substantial—i adjoinind en e e seeAntonine th b gth t fign na i n 4 .ca t e Vallum. On the south side a very small trench, bedded with stones, was noted running east and west, under the rampart, about 10 feet back from the lip of the ditch. Its direction rather negatived the idea that it had bee samshalloe drainager o th nfo to t es a timwd wa fot ean , i palisadera - trench. Possibly it was dug to receive the lower ends of struts which had passed at an angle through the body of the rampart and had supported a facing of wooden planks. An arrangement of the kind was note t Ursprinda Germae th n go n Limes.2 firse th t n cross-sectioI madee nw ditc e abous th , hfeewa 5 1 t t fee9 broa d e t deeth dan n i p ; next1 , 6 feet broad and 3 feet deep; in the third, which was 8 feet distant from the kerb of the Vallum, 4J feet broad and 2J feet deep. I did not myself see the last two of these sections. Mr Smith tells me that to him they suggested half-finished work. Pabriciue Se O.R.L.,n si . vi. , KastelBd B , l Urspring (Nr. 66a). 7 . ,p 2 254 PROCEEDING E SOCIETYTH F O S , MARC , 1933H13 . A swere longw es a explorin e outegth r fringee annexeth f o e s w , were in comparatively smooth water. The next stage of our enquiry fas r wa more troublesome facquestiono e t ed Th wer. ha e esw extremely complicated, and the obstacles in the way were numerous and formidable. Some of the problems that presented themselves proved impossible of solution without the removal of trees and a much more extensive disturbance of the soil than we had either the means or the authority to undertake. This applies more particularly to what may well have been trace Agricolaf so regards A . Antonine sth e othe fortth n ro , hande w , t onl no o obtainyt t securet alsou bu , o t e informatiose th d d ha e w n stumbled upon facts which were as new as they were unexpected, and which illuminate its history in the most interesting manner. The feature e plath nf o swhic h have hitherto seemed unintelligible wile b l found to fall quite readily into place. On enterin e annexeth g e begaw , n operation n thai s t t i par f o t which lies just outsid e north-easth e t e cornefortth ,f o betweer e nth Military Way and the Antonine Vallum. Even a casual inspection of the original plan (fig. 4) makes it evident that the 'lay-out' here was peculiar. The absence of any ditches in front of the rampart of the fort and the appearance of a, rectangular ditch some distance to the east of it are very remarkable features, on which it seemed eminently desirable that fresh light should, if possible, be got. We thought it well to make sure, in the first place, that the fort ditches were really absent. There wer o visibln e e signe surface f t th thei o s t n mo bu , was just conceivable that they might nevertheless be lurking beneath the cobbling that had been discovered in 1903. This cobbling, which extended over practicall e e enclosurwholth th y f o e n whico er ou h attentioconcentratedw no s nwa s coverewa , d wit hspreaa f graveldo , d consistean f freestono d e rubble, closel d carefullyan y laid upoe th n natural surface, to a depth (as it seemed) of 14 inches. Lifting it for e breadtth a cutting f ho , made transversely acros e linth se whice hth ditches would have followed, we found that the till was nowhere disturbed. Mr Buchanan and his colleagues were, therefore, right in believing that there had never been any ditches there at all. The enclosur d bee ha n eintegraa n e original th par f o t l e desigth f o n Antonine fort. Speaking of the rectangular ditch, the Report of 1905 says: "To the interior .. a .rampar s stili t l traceable, which take a e for f th smo prominent mound at the south-east corner. In the centre is a core of stones, among whic s observewa h d evidenc e effectth f f fireo o es . The soil forming the rampart is not laminated." If the enclosure belonge e Antoninth rampare o dt th d e di periodto thato o ts , defended ROMAN FORTS AT ROUGH CASTLE AND WESTERWOOD. 255 thoughe w d whethee t an welse i t , o it coult le w r d mak descriptioe eth n of it more definite, particularly as our curiosity had been whetted by the reference to "the effects of fire." Fig. 4 shows that the section on whic e earliehth re e wes accounth a littl th f t o to et cu base s i ts dwa corner—that is, in a north to south direction. We decided to keep clear of the mound, which is still conspicuous, in order that we might get to the bottom of things without having to move a mass of soil. Accordingly r trenc ou poina t t fee6 ha cu 2 t t e e cornernortw , th f ho — that is, in an east to west direction. The result of our observations (fig. 6) may be summarised as follows:— The freestone cobbling stopped short, with a well-marked edge, about 12^ feet west of the lip of the ditch. The surface beyond it offered such an obstinate resistance to the spade that we were at first doubtful whethe e tillth . e t mighi b rPerseveranc t no t e proved that

Fig . Sectio.6 Cobblef o t Easna d En td Enclosure. it was the upcast from the ditch, with its upper inch or two caked to a consistenc f singulao y r firmness . a "sigo corn Therf o nf s o e wa e stones." But, on clearing away the ' forced' soil below the crust, we found thae layerth tf a upcasmaximuo d ha t m thicknes f abouo s t 2 feet and that towards the east there was intermingled with it a mass of earth, reddened through and through by fire. Beneath the upcas beginnind an t a distanc t feeg6 a, 1 tf eo wes t fro e centrmth f o e the ditch and 7 or 8 feet east of the edge of the layer of cobbles, there was a second layer which extended westwards under the first. It was lesno s carefully laid, but, instea beindof g composed entirel lumpyof s of broken freestone, it contained a large proportion of stones which d evidentlha y been gathered fro surfacee mth distanca t A . f 3eo J feet in front of this lower cobbling, but still buried beneath the crust, was quita e unmistakable post-hole. Wit viee h th f ascertaininw o g whether it had been one of a series, we enlarged our cutting for a short distance, first to the north of it and then to the south. In doing so, we encountered in both directions, at a distance of about 3J feet from its centre, a shallow circular depression among stones, with a little black matter (which might have been decayed e woodbottomth n i ) . These depressions were certainly not post-holes in the full sense of the 256 PROCEEDINGS OF THE SOCIETY, MARCH 13, 1933. word. But it is not improbable that they had held the lower ends of wooden struts which ensured the stability of the post that rose between them. As it turned out, this first cutting, taken along with the remains of the rampart shown in fig. 4, gave us all the elements which were essentia reconstructioa r fo l e histore cobbleth th f o f nyo d areat Bu . e informatioth e verifieb t d supplementei o yieldent an dd ha d y b d

ANTONINE RAMPART

LATER COBBLING SUPERIMPOSED UPON EARLIER §: 13 PIT7

=- il HEARTH all PH=r. ~

MILITARY WAY

10 O 50 IOOFEET h

Fig. 7. The Cobbled Enclosure. cuttings made elsewhere, before the interpretation embodied in fig. 7 bega assumo t n e intelligible shape. Although hertherd an e e parf o t n outlina remaiy ma e n doubtful, e illustratioth 1 acceptee b y s nma a d a reasonably accurate record of the essential facts. As a record, however, it does not profess to be complete, for to have made it so would have calle r mucdfo h more tim word ean k tha coule nw d afford o devot. t it Thus o refrainee t ew , d from attemptin o locatt g e th e whole of the post-holes in the series that must have run along the east front. When another cutting, close to the corner, had disclosed a second post-hole in exact alinement with the first, we were content to tak e existenceth e resr grantedth fo tf eo thir A s discovere. dwa t a d the same time, not (as will be seen from fig. 7) in the same alinement otherse asth t 9| bu , feet south-west-by-wes e secondth s sizf It o te . This is particularly true of the south-east corner, which had been much disturbed in 1903. 1 ROMAN FORT T ROUGA S H CASTL WESTERWOODD AN E 7 25 . and depth indicate dheld exceptionalln thaha da t i t y stou heavd an t y post, in all likelihood the first of the series. The limits of the two layer s f cobblingthea o s r y fa wer s a , e definitely ascertainableo s e ,ar plainly marked in the illustration that it is unnecessary to describe them in detail. Attention should, however, be drawn to the paving which began about 11£ feet west of the large post-hole and covered a space measuring, roughly . 9 Ther fee3 y 1 ,b te wero layertw e f o s , jusit s thera t ee cobbling werth f eo thesd an , e were separated from one another by a foot or so of upcast, part of which was reddened by intense e heatsign Th f firo s. e seeme o t stod p here after being detected at intervals, in front of the lower stratum of cobbles, all the way from the Antonine Vallum. The quantity of the burnt material varied in the different sections, and occasionally some of it lay on the top of the cobble-stones. Befor attempy an e mads i t interpreo et t these various appearances, it is necessary to return to the west end of the area and pick up a clue which we were fortunate enough to light upon there. As a first step, however t shouli , pointee db t tha representatioe dou tth rectange th f no - ular ditc fign incompleths i i .4 inaccurated ean , althoug t hcorrespondi s fairly closel seee whab o yt nn upoca t surfacee nth actuae Th . l dimensions are as set out in fig. 7. Here it is necessary to anticipate a little, and to explain that the fort rampart as it appears to-day is not the fort rampar t i existe s a t d whe e ditcs firsth n wa ht dug e shalW . l learn presently that an addition of no less than 9 feet was made to its width on the outer or east side, when the castellum was reconstructed for the second timeReporte Th . , whic hers hi e more explicit than fig , fixe.4 s the starting-point of the ditch "about 20 feet in front of the rampart"— tha really, tis addition e fronn ,i th f o t . Thi eastr fulls fa si fee 6 o .y2 to t It actually started about 3 feet out from the kerb of the original rampart, so that some 6 feet of it must be buried beneath the extension. At first it is comparatively small, not more than 5 feet broad and 3| or 4 feet deep. A t advancei s s eastwards, however s dimensionit , s increase. Befort i e reaches the turn, it is quite 14 feet broad, with a proportionate depth. After rounding the corner, it has a uniform breadth of about 17 feet, unti t stopi l e Antonin fronn th i s f o t e Vallum. The first proof that any amendment of the plan of 1905 was called for obtaines wa course th n df testini eo ghypothesia s whic beed hha n very tentatively advance 1911.n di With nothinmateriae th t gbu l available

in the Report to go uponl , I had suggested that the ditch, which had an odd appearance of being intrusive, might really be an adaptation of some 1 Roman Wall Scotlandn i (ed. 1Q11) . 235p , , where "north-east" imisprina s r "northfo t - west." VOL. LXV1I. 17 258 PROCEEDINGS OF THE SOCIETY, MARCH 18, 1933. older—than a par f o t t n isAgricolan—systema , f I thi. s were so,e -th original ditch must have continued farther westward. A transverse s accordinglcuwa t y mad greao n e t acrosta e apparend th san p ga t distance e rampartfro e edgth m th f eo . Whe e coverinnth g soil was stripped offcobbline th , sees havgo wa nt e subside l alondal edge gth f eo Militare th y Way; ther certainld eha y bee nditca htimee theron .t ea On removing the cobbles we found that it had been V-shaped and that it had been deliberately filled in. So far the omens had been favourable, but disillusionment came speedily. There was only an inch bottome th silf t o thed a t o an , ntw camr o feee2 earthf o t , mingled with large stones and broken pottery. Higher up, the filling was almost entirel f lumpyo f brokeo s n freestone f whico , h there wer layero etw s with a thin spread of gravel between. The lumps in the upper layer were of greater size, some of them looking as if they had previously done whole dutsecona th s building-stoness yf a eo wa d p spreato e f th do n O . gravel. We were plainly in the presence of two roads, one superimposed upo othere nth t waI . s evident thaentrance th enclosurte th o et d eha been here, and that the traffic had been heavy. The negligible quantit e ditcth f silhyo n i t tende o discredit d e th t Agricolan hypothesis. The testimony of the pottery fragments was still more discouraging. The majority were portions of amphorae, too much mutilated to convey any meaning. But there were also some pieces of Samian ware, including the bottom of a platter of Form Dr. 18/31 and e greateth r a smalpar e f decoratioo t Th l bow. f 37 Foro f l. o n mDr lattee th r is characterize pooy db r workmanship stampe th ; wore sar n and carelessly impressed, and the ovolo is singularly crude. Its surface, too mucs i , h rubbedJamer M t s bu ,Curie wa o firse examinsth o wh , t t , eit recognized the following types immediately—Dech. Nos. 187, 265, 455, 613, and 1041. Two others have since been identified—Nos. 471 and 555. There remain only a small leaf-like object, which does not seem to find a plac Dechelette'n ei s lis t d threalla t an , e figures f eaco ,f whico h h no more than a part, in two cases a very small part, has survived. Now it cannot be a mere coincidence that every one of the seven stamps enumerated is known to have been used by Libertus of Lezoux. Dr Oswald suggests Vespasian-Trajan for the floruit of this potter. Others believ s stile wa ltha e activh t e under Hadrian t laterno f .i , trute Whateveth he b athato st y Roug e th ,rma h Castle bowl, wits hit degenerate craftsmanship, cannot possibly have reached Scotland earlier tha e Antoninnth hardln eca periodt i y d havan , e been thrown upon the scrap-heap there as soon as it arrived. Antonine troops had therefore been in occupation of the site for some time before the filling took place. Agricola finally faded out of the picture when it was discovered ROMAN FORTS AT ROUGH CASTLE AND WESTERWOOD. 259 that the ditch ended 3 feet short of the kerb of the original rampart, were fow r e abl satisfo et d yan trua ourselved s een wa d sen tha e th t not a break for a gateway. The rectangular ditch, then, had belonged to the initial lay-out of the Antonine fort. When the first reconstruction took placebeed ha n t ,i fille witp du h debriearliese th f so t Antonine occupation roaa d d an ,carrie d acros t intsi enclosuree oth . Whee nth a secontim r efo d reconstruction came e roacollapsed th , dha r fa o s d necessars thawa t i t replacdons o thid yt simply wa an seb , eit y laying a second road upon the top of it. Other roads in and about the fort undergond ha e similar repair, notably tha toriginae parth f o t l Military whicy Wa h extended fro ease mth e tcastellum gatth e easf eo th t o t gate of the innermost and latest annexe. The double stratum was unmistakable there, and it occurred to me that it offered a convenient touchston e soundnesth f e theoro e advanced th th ha f o so yI t s da chronological sequence of the various annexes. The road \vas, therefore, examined at a point a little way inside the central annexe-ditch (fig. 4). It showed no such signs of repair as had been noted farther west, a very satisfactory confirmation of the view that the area beyond innermose th t ditc beed hha n excluded fro annexe openine mth th t ea g e three thirth oth f e do Antonine periods face Th t .tha gata t e still led int t osuggesti t entirelno s s ythawa abandonedt i t . Perhape th s remains of its eastern rampart were considerable enough to admit of cattler fo n servins .it pe a s ga e discoverTh y tha e roabeed on tdha n superimposed upon another prompted us to scrutinize the cobbling of the interior more carefully. t I turne t thaou d t here also there werlayerso tw e , separated from one anothe thia y nb r sprea f gravedo eacd an lh abou inche7 t s thick. A cutting carried westwards from the apparent edge of the enclosure supplied convincing proo f e sectionthiso f Th . s reproducea , fign di , .8 s self-explanatoryi t attentiobu , n mus e drawb t o certait n n features ot whici f h hav ewidea r bearing cuttine Th s continue. gwa r somdfo e distance e intrampar t wil i e bod th e fortseee o d th b f th l yo nan f , o t that the lower cobbling marched with the original kerb as closely as it would be possible to do. The upper cobbling, on the other hand, besides being higher tha kerbe sunderens th wa , intervan da froy b mt i l cuttine th side f th e laminatioeo feet 9 gfth n I . s vernwa y distinct abov stone eth e foundation blace th , k lines tendin drooo gt plittla s ea they came directly over the kerb. But it was no whit less distinct feeabov 9 f lowe o e t th e r cobbling d her e blacan ,eth k lines wer- ein cline riso dt s the a e y lefe kerth t b behind them runo . Evewh s e nh y readma . Whe e cobblinnth s firsgwa t laid down e outeth , r facf o e rampare th t remained unchanged. s renewedWhewa t i e noute th , r 260 PROCEEDINGS OF THE SOCIETY, MARCH 13, 1933. e ramparth fac f s feet give9 ek' o en othe I f ewa n to . a 'n r words, the final reconstruction of the fort had involved a reinforcement of its defences shale W l. retur thio nt s shortlye meantimth n I . t i wile l be best to complete the account of the annexe. In one vitally important respect the evidence was still defective. t knono wWd di ewha t relation f anyi , e doublth , e laye f cobblino r g on the west had borne to the double layer we had previously en- countere e east t firsth A .n t o dsigh t there a seriou seemee b o st d y wa f supposiny o beed an y ha n ni wa o obstacl ge tw th tha e n i eth t connected. In the former case the two layers had only a sprinkling of gravel betwee nlattee themth rn I the. y were separated froe mon another by nearly 2 feet of soil. Investigation, however, proved that

I O lOFEET

BASE OF FORT RAMPART Fig. Sectio8 . f Cobble o t Wesn a d En td Enclosure. the intervening soil had all the characteristics peculiar to the upcast which had already given us trouble, notably the tendency to cake, and, further, that it decreased steadily in thickness as it was followed westwards. The levels were then taken, with the result that the height of the lower cobbling above the datum line was found to be virtually the same at the east end of the enclosure as at the west. There could no longer be any doubt as to the double layer being continuous throughout e upcasTh . t d thamadha s thesitat u e e meano n t more than that the ditch had been deepened immediately before the proces f re-cobblino s g began. Whe e finath n l reconstruction took place defencee ,th enclosur e beeth d f ha so n o strengthenedeto . e e beewholevidenc w t th out Th conclusionno e nse f o es Th . ha e s whico t t seemhi statee b leao n st ddca most clearl narrativn yi e form. The enclosure, which measured roughly about 130 feet by 60, was an integra Antonine l parth f o t e for originalls a t y designed thes nwa t I . e eas th ditca t n o e sout hth d whicn han uncobbledo d h seemha d an s, to have been about 5 feet wide, and which was apparently a line of demarcation rather tha na barrier positioe e entrancTh th . f o n s i e quite uncertain. Equall o whict ye reservee uncertaith hus e th d s i n ROMAN FORTS AT ROUGH CASTLE AND WESTERWOOD. 261 confidentle putb e l thab n o Al .t ca t ares yawa affirmes thas wa i d t i t intende o supplement d e verth t y limited accommodation available within the ramparts by providing room for something that was nor- mally placed inside—perhaps a barrack-yard for drill. The first of the two reconstructions which the castellum is known to have undergone was accompanied by three notable changes in the enclosure. Cobb- s linlaidwa g down ovee wholth r e surface a roadwa, r wheelefo y d traffic was carried across the ditch, and a timber barricade was erected all along the eastern front from the Antonine Vallum to the corner. barricade Th s apparentlewa y compose f horizontao d l planking, nailed firmly to stout posts or valli which were set up at intervals of perhaps feet0 d eac1 f an whicr ,ho 9o s probabl hwa y supporte eithen do r side a woode y b n strutmannee Th . n whici r e cobblinth h g spreadt ou s towards the ditch at the south-east corner may indicate that there entrancn a s r foot-passengerwa efo s here, whil pavine eth repre y gma - sent the floor of a small building, possibly a wooden hut. Modifications so sweeping ammint to more than repair. They indi- cat e adaptatioeth e enclosurth f no o somet e quitpurposew ene e W . shal course l du fin n ei e dver thatth yt a ,tim e whe e alterationth n s were made, the space within the ramparts was cut down, and cut down, as it happens, by approximately the same number of square feet as the enclosure contains. It is reasonable to infer that there was a transference. Stores—barrels of wine, for instance, and amphorse—may now have been kept outside, although sufficiently near the gate of the fort to allow of their being moved inside quickly, if there were a prospect of the garrison being hard bestead. A serious defence of the relatively flimsy barricade can scarcely have been contemplated, particularly in the absence of any corresponding protection along the south front Why, then, hav ebarricada t allea ? Probabl precautioa s ya n against the entry of unauthorized intruders from the annexe; it is hardly to be supposed that 'broachin e admiral counterparo th gn d ha ' t among the Romans. Nothing of the sort was needed on the south, since the line of the road, from which alone access could be had, would be under the immediate eye of the sentries at the gate. Such would seem to be e easiesth mosd an t t natural explanatio adjuncn a f no r t fa whic o s , his as I remember, unique among Roman castella at home and abroad. Its occurrence here e a abnormallmusresulth e f b to t y small sizf o e e fort—lesth s tha n acra n e aftee reductioth r o whict n I hhav e referred. When the Romans withdrew for the second time, the barricade was given over to the flames, whether by the retreating troops or by the exultant Caledonians no one can tell. The stumps of the valli and of 262 PROCEEDING SE SOCIETYOFTH , MARC , 1933H13 . their supports were left unscathed below the surface, to perish later by natural decay. Everything above blazed furiously, and, crashino gt groundthe , burned itsel thereout f , reddenin soiwhicgthe on l layhit . Before very long the invaders were back at Rough Castle once more. When they set themselves to raise up the former desolation they d lines t ol witadhere bu e , hth certaio t d t unimportanno n t differences (fig. 7). To begin with, the ditch bounding the enclosure was widened and deepened, the increase on the south side being gradual and progressive, that on the east uniform and at the maximum. In e procesth upcaste s th mucht allf o no , ,f i ,includin reddenee gth d earth, was thrown inwards on to and above the original cobbling and post- holes. Fresh cobbling was then laid down on the new surface, as well as on that part of the old surface which had not been reached by the upcast, and on the roadway over the ditch. At the same time the buried patc f pavino h s replacegwa anothey b d f virtuallo r y identical dimensions. As recobbled, however, the area at both ends was shorter fee9 r tbeed o b thay8 ha n t ni before . Fig make8 . s cleathiy swh r e e westcurtailmenreasoe th th Th r n . e easo s nwafo th o wa ts n o t not dissimilar. An earthen rampart, the remains of which are still visible, was substituted for the wooden barricade, and the necessary room for its base could only be found by taking something away from the cobbling.

histore e enclosureTh th f yo , then, falls wholly withi e Antoninnth e period. But, in collecting the evidence which has made its re-discovery possible e camw , e upon trace f somethino s g which e seemcerb o -t s tainly older. Thus, in cutting a trench east and west along the Berm e Antoninoth f e Vallum e strucw , kvera y small ditch fee4 , t widd ean about 1J feet deep, running north and south, on a line parallel to the e cobbleditcth f ho d enclosur a littl t eebu farthee easth t o t (figr. 7) . Toward s sooe nort greae wa th Ditcse n th t th hi losf tho n i tWall n o , the farther side of which it failed to re-appear. Towards the south, othee onth r hand t passe,i d unde stone th r eVallume basth f eo , where it was transformed in a manner which proved it to be unmistakably pre-Antonine. Beneat e nortth h hd bee kerha n t bi fille d with stones to carry the later structure; as it approached the centre it grew shallowe shallowed an r hardls r untiwa t yi l visibl t alla e ; thereafter e procesth s reversed wa se tim e th souto tha th es y , b ht kers wa b reache returned ha s origina t it di o dt s oncl wa depte d morhan e filled with stones. It was obviously older than Lollius Urbicus. When the Wall-builders arrived upon the scene, they must have found it crossing a slight rising ground, directly in the path marked out for them, and ROMAN FORTS AT ROUGH CASTLE AND WESTER WOOD. 263

must have removed part of it in the course of providing a level bed for the foundation they had to lay. That the little ditch had continued beyond the south kerb was apparent, but the intervention of trees made any attempt to follow the trail hopeless. Fragments of similar ditches—for they can hardly have been the same—were found farther south, in positions which are marked on fig. 7, the most noteworthy being a rounded corner. There alss ohearth—whicwa a quity hma e well have been Antonine, although it was a foot or two below the modern surface—and what seemed to have bee roundee npita Th . d corner would have been more thoroughly examined, had not progress been once again completely blocked by trees. That it had had no place in any Antonine system of fortification was clear from its having been deliberately choked with heavy boulders. None of these had been used for building. They had been gathered for e expresth s purpos f blockino e e ditchgth . Eve combination ni e nth indications just described are far too slight and vague to justify any conclusion except one of the most general character: should it ever become practicable to initiate an exhaustive search for first-century remains explorere ,th s wil t leasa l t know where they ough begino t .

Uthio pt beed s poinha n e workinw t g entirel annexee th n yi . Inci- dentally, however, we had discovered that, when the fort was reconstructed for the second and last time, the rampart on the north-east had been extended outwards by 9 feet, and there was a prospect that r handwit ou mighe n hw si e ablethiy b t ke s, without much trouble, to add substantially to what the Report of 1905 had said of the defences as a whole. So far as the rampart-base is concerned, the conclusion there reache s summei d s followsaffirmee a b p y u d ma d: "t I that underneat e rampart th e for l th hal tf o thers e stonear e foundations of an average width of not less than 20 feet, supplemented by varying margins, adapte o suidt t special requirements increasind an , g e widtth nowherthas o i hs t i t e less tha feet 0 amountd n3 an e ; th n i s eas superstructure t th rampar o t feet. 5 s 3 A o "t tole t ar f turfd eo e w , that "whil notes i t ei d thae layerth t s terminate e edgth th f o et a e stone foundation, similar laminated soil appears beyon s kerbsdit , both externally and internally, lying on the original surface,1 extending out- wardly 6 feet and inwardly fully 8 feet, sometimes in continuation of e core t th morbu , e often quite separate alwayd an , s showin e samgth e systematic layering, evidently intentionally laid, to add to the width of the rampart." The phrase "lying on the original surface" is hardly borne out by the sections in Pro- ceedings,1 vol. xxxix. (1904-5), majoritye pi . th i . Some t no f i ,, sho wbottomina stonef go . 264 PROCEEDINGS OF THE SOCIETY, MARCH 13, 1933. The accompanying sections prove that the estimate of 20 feet as an average breadt e originath th r f efo o h l foundatio y quitma ne safely be accepted. The idea of " varying margins," however, is a priori unsatisfactory uniforA . m r mor fa breadte whol th a e r s likeli ehfo y arrangement. The 9 feet of fig. 8 gave us a standard for the outer n examinin('o margin' e eastd th an ,e n Repor o th g t i turnet t ou d that,1 in the ,s,ectipns cut in 1903, the outer 'margin' was exactly 9 feet wide on; the; west \and. only a very little more on the south. A section cut; onthej east, but-to the south of the Military -Way; had given a rather different result, the "margin' there becoming "a strong stone revetment, about 7 feet 6 inches wide, built higher where facing the •interior, and set back 5 feet from the edge of the inner trench." It shoule .lookew f i d s dhava reckoo t e n with variety after all. Before • acquiescing matte, e tese howevermakinth y th t b t o rt pu e gcuttina w , g describee on r froReport e e fa th 'margin me th t n di of,ouno Th .n ow r' whic t disclosehi s certainldwa t lesyno s fee 9 tha e t d nth whic ha e hw standarde thus .assumei th t i se b scarceld o an dt , y doubtful that.the •deficiency of-l| feet, whicr predeces'sors s noteou wa hy b d , -must,.be due e accidenta.toth r ltw o remova oe stoneson f o t lisI ., therefore, fairly safe to say that on all three sides the breadth of the extension had .Originally been about 9 feet. • ' ' • ; ;U The Antonine Vallum along the front of the fort would a,ppear:dfo have been similarly reinforced, presumably at the same time. There, ^however, the sods seem to have been laid on the natural surface without .any.- bottoming of stones, and the precise breadth of the extension! -is .consequently difficul determineo t t the" as y r .e facts fa th Her o e s , ear .have been observed. Speakin e Antoninth f go e Ditch Repore ,th t says : bere ;"Th m nex fore tth t is abou fee7 2 t t wide, measurin stone th o et p gu .foundatione raimparth s a t t bu extend; s beyon e foundationth d e th , width is reduce abouo dt feet.0 2 t " This implie breadtsa abouf ho feet7 t e actuafoth r l extension Glasgoe Th . w Committee, againsingle th n e,i •section which they cut here ten years earlier, noted that lamination was well marke distanca r dfeefo t fro6 stone edgf e ou teth mo th f eeo base. But they also mention that 6 feet farther out still there was "a stone "kerb."e a quasi-kerb,e th whic b f I y l hma takee thes b " a o nt y s cali , it l as the limit, it would mean an extension of 12. feet. The figure may not seem excessive rememberee b t i f i , de sam thath en i tsectio stone nth e foundation proper was only 16 feet wide, as against 20 in the case of the ramparts of the fort. Finally, at this particular point, which was just north-wese easth f o t t corneforte foundatioe th th ,f ro sais ni havo dt e lacke neatle dth y formed kerbs which occurred everywhere else. Instead, 1 Antonine Wall. f Report,6 11 . pp ROMAN FORTS AT ROUGH CASTLE AND WESTERWOOD. 265 there was "at each side a built-up core of stone about 2 feet high"— 3 feet wide on the north, and on the south 2 feet 6 inches. The suggestion of repai unmistakables ri likelihool repaie al th n i d associates e ri b ,an do t d wit firse hth t reconstruction, sinc extensioe eth beyony nla . dit innee th margin,o ' r t I comw eno ' wher e difficultieeth f interpreo s - tatio more nar e serious. Speakin ease th tf gsideo Repore ,th t states that towards the interior, beyond the kerb of the foundation, "there is a channel, stone-bottomed inche6 , s fee6 dee y t pb wide . This, agains i , bounded inwards by a raised core of stones of about 3 feet in width, which ends upon the edge of a cobble-paved street rising about 9 inches abov . eit This.stree fee2 1 ts i twide. descriptioe Th " bases ni d upoo ntw sections, cut respectively to the north and to the south of the Military Waye drawin e latteTh th . f ro g break f abruptlof s y just whers it e testimony would have been specially valuable e forme t th tha f bu , o tr shows both the " channel" and the " raised core," with the layers of turf e " street.e edgth coverinth f eo s a " r whol e fa gExcepth s ea t e foth r differenc e characteth e stonn i th e f eo r bottoming, wha seee w to m t have extension hera s e rampartinside i th th f f eo no , very similao t r that upon the outside. The street would then be exactly in its right place in the intervallum. Our own evidence from the east was less definite. When we were working there, we were chiefly concerned wit outee hth less r sinne wa e face thoroughle Th ron . y exploredl al t bu , appearancee th s pointe thero dt e having beedistinco ntw t layer stonf so e ' bottoming,' stretching as far as the original rampart and having laminated soil between them. On the west, according to the Report, the inner ' margin' had " flat stones like pavin top layere e seee th b g Th .nturn f n " so o spreadin ca f g inwards over these both in the drawing of the section cut in 1903 and in the contemporary photograph (fig. 9). What the Glasgow Committee have to tell us is even more conclusive. Their section through the west rampart "begins nea e easterth r n kerbston f Avhao e t appear o havt s e beea n roadwa fee4 ywidth1 n i t , paved with flagstones, thre whicf eo h measure abou fee 2 tly ib t feet inchesinche4 2 3 2 , d 19y b s1an , 1 inche19 y b s inches. It passes through a heap of confused stone, mostly squared, and indiscriminately thrown together, intermixed with loose earth. This covers about 8 feet of the line of section. Then begins the familiar layering, as seen in all the Bonnymuir sections. Through this the section proceeds, and just after fairly passing through it, ends on the edge of the inmose westerth f o t n ditche e camp—ath f o s poina t fee2 5 t t froe mth eastern e supposekerth f bo d roadway." *• They add that, while the confused heap of stone and earth appeared far too loose to be the remains of buildings in situ, the stones are dressed 1 Antonine Wall Report, . 118p .

ROMAN FORTS AT ROUGH CASTLE AND WESTERWOOD. 267 yellow freestone furthed an , r thapoina t feea 8 t 1 t t fro beginnine mth g of the section, the fragments of a large vessel of earthenware were found nearly 3 feet below the level of the pavement. The suggestion of a roadwa hardlys i y probable. Streets within forts wer t usualleno y paved, a castellum n i d smalo an s s Rouga l h Castle there would scarcele yb roo14-feea r mfo t stree than i t t particular position courtyarA . e th r do floor of some building is more likely. But, for our present purpose, the important thin tha s gCommittee'i e th t s statement t standsi s ,a , gives su a continuous belt of 44 feet of laminated soil. If we deduct 20 for the breadth of the original rampart, we have still 24 to divide between the extensions fee5 , r twhico mor4 s hi e thae Com requiree th n w d - Ha . mittee given complete details of the stone bottoming, we might have been able to dispose of the surplus. As it is, they deliberately confine their observations to the base of the rampart proper, so that we must leave things where they are, remarking only that the spot where the amphora fragments were found must have been easjuso footw a t r o t of the original kerb, which means that they had been used for making e grounupth d whe e extensioth n addeds nwa . Our own experience on the west had one or two features of special interest. The stone bottoming bore some resemblance to that which the Report describes as characteristic of the east. A bed of stone pitching extended inwards from the kerb for 7^ feet, when it abutted agains somewhata t higher ban f stonesdo fee,2 inche9 t s wide. Beneath the latter was gravel, arid above it was earth, reddened by fire. The soil ove e stonth r e pitchin richls gwa y laminated laminatioe th , n being interrupted at a fairly low level by a layer of large stones. We pushed our section far enough west to uncover the junction between the superstructure of the extension and that of the original rampart. They were standing side by side to a height of 3 feet 10 inches above the kerb. The line of demarcation, which was extraordinarily distinct (fig. 10),J indicated thae loweth t originae r th par f o t l rampart-fac begud ha e n to rise at an angle of about 72 degrees and had then developed an inclination inwards. With a base of 20 feet, the batter suggested would easily admit of a height of 10 or even 12 feet. Anything more than that would probably hava platfor s a e reducep mto e value th dth f eo for defence; the object of the extensions would be to increase the breadth. The gravel underlying the band of stones may represent the remain e earliesth f o s t intervallum street, which must hav n closru e e a The photograph [was taken for me by Mr Smith. Although the layerings in the extension are barely discernible in the illustration, they were in fact very pronounced. The difference in the degree of response which the two sets of sods have made to the camera indicates a difference in the character of the soil from which they have been cut—in itself a proof of difference of period. 268 PROCEEDING E SOCIETYTH F O S , MARC , 1933H13 . to the original kerb. The significance of the burnt earth is more obscuret occurrei d Ha . d nowhere else I ,shoul d have been inclineo dt associate it with a hearth, for in the Vallum forts hearths are not infrequentl e westerth f o ne le ramparty e founth n i d, which afforded ready shelter from the prevailing wind. But we met it again both on southe I canno th d n an ,o norte t d th helhan p wondering whethee w r ought not rather to think that a flight of \vooden steps, giving access e rampartth f o s bee p ha , nto destroyee th o t firy b de whe e forth nt

Fig. Sectio10 . n throug Wese hth t Rampart, looking south, showin e sode th g th f so extension abutting' agains originae th t l inner face. abandoneds wa . Unles precepte sth f Hyginusso were entirely ignored,

a whole series of such flights l must have risen from the intervallum street. The Glasgow Committe y nothin sa o ewhat s ga t they havma ye found on the inside of the Antonine Eampart, and the excavators of 1903 are equally silent. We carried a trench from its southern face backwards into the fort for 10 or 12 feet. At the lowest level and close to the kerb were stones and gravel, doubtless representing all that was left of the original intervallum street. Above that, the soil immediately adjoinin Vallue gth laminates mwa heigha inches 5 o 1 dt f o t , but the lamination rapidly tailed off. On the top of it was earth reddene firey db . Then cam elittla e more lamination, less well defined, and finally soil and stones. At a distance of 9| feet out from the kerb were the remains of two roadways, one above the other, presumably 1 Meminisse oportet in hostico ascensus valli duplices et frequentes facere (De Mun. Castr., 58). ROMAN FORT T ROUGA S H CASTL WESTERWOODD AN E 9 26 .

the intervallum streets of the second and third periods respectively. Th d beee ha lowen e practicallon r e samth en e streeo yleveth s a tl that lay between it and the kerb. As might have been expected, the surfacinterioe beefore th d th n ha tei nf o rhighe r durin perioe gth f do the last occupation than during that which had preceded it. The south rampart had not been touched by the Glasgow Committee. n 1903I , however t righa sectio,cu s t througndescriptiowa o N . hit n resule oth f t Reportappeare drawine texe th th f th t o t n bu i s, g shows layers of turf on the inside, apparently resting on the natural surface extendind an g backward r 6Jsfo feet, after which come stones wito hn layer f turo s f above them e conditionTh . t sa wit whicme n i h e w h e samth en trenc o t sid farthet bu ecu h r east were quite different. Close to the rampart-kerb, and on the same level, we found a layer of e absencstones t possibl th positively no n sa I f graves e.o o et wa t i ly that beethid earliese ha snth t intervallum street f courset thao , bu , is t , wher e woulon e d have . lookeit Restine a stoner th s fo d n wa o sg laminated stratum inche6 , s deep, which coul e followeb d fee 8 r t fo d northwards. Over that, again, there seemed to be a second layer of stones, intermingled with gravel, 1J feet thic t graduallkbu y diminish- abouo t g distanca foo1 in tt o a tfee9 tw f tr e o o fro kerbe e mth On . e topmosoth f t stones were laid flat like paving-stones, recalline th g description of the west side given in the Report. Above the layer was more lamination, as well as earth reddened by fire. This upper stratu abous mwa foo 1 t t maximume thicth t kt a couli d tracee db an , d for only some 3 feet inwards from the kerb. e evidencAth s e regardin e insid th s gles wa se unambiguous than could have been wished, I have set it out at length. Everyone may t agreno e wit e conclusioth h t I nseemhavi e em s reachedo t t Bu . certain that inside the fort, just as outside of it, the rampart was 9 occupatioe th f o d fee0 oe 1 n en rt beed th thae widet ha nth a t ni t a r beginning. Two questions immediately arise. What was the object e firstth ,o t e additiocarefu t i s omadth d fwhe A s An e? wa ln n ? consideratio t asid se idee eth o t a e tha extensioe nm th thad le s d nha been mean o support t raisea t d walk, oftes suci s nha found, bott ha home and abroad, immediately behind the walls of stone forts erected in the second century or later. The superstructure has, I believe, been e samoth f e e ramparheighth s a t t agains t s rearedwhicwa t t i hi ; serve reinforcemena s da t pursimpled ean e additio, th exactl d di ns ya outee thamads th raisea e ide wa f tth o r e t ao t faced Bu wal. k suggests some points of so much importance for the Vallum forts as a whole that I propose to discuss it separately in a Postscript. Mean- whil secone eth d answered e questiob o t s nha . 270 PROCEEDINGS OF THE SOCIETY, MARCH 13, 1933. Before we began our examination of the extension on the inside, I was quite prepare accepo dt viee th tw thatt weri f extensionn i ,e a t ,i had been made at the same time as the one on the outside. What we actually found, however, led me to change my mind. Despite the differences in detail, we seemed everywhere to be confronted by at leaslayero tw tf stones, o s 1 each wit a laminateh d stratum abov, it e an plausiblo I dn coul e se d e explanation excep supposo t t e thae w t were dealing with the debris of two extensions, one of which had been othere ruine th th heape f n so . o Bothp du were foreigoriginae th o nt l scheme. The lower must, therefore, have belonged to the first of the two reconstructions of the fort. The upper can only have belonged to the second, and we already know that it is with the second that the extension on the outside was connected. Nor is it at all surprising that the lower one should have become so dilapidated as to require rebuilding, thad an t to-day only miserable remnant uppee shoulth e f sron o d survive. It was the inner face of the rampart, with its numerous ascensus valli, that would be most likely to attract the attention of victors bent upon destruction. Shor f levellino t e defencesgth e sures,th thas t wa t way of rendering them useless.2 A review of the evidence from the cobbled enclosure lent additional suppor chronologicae th o t t l sequence thas jusha tt been proposedf I . the first of the two additions on the inside were assigned to the opening of the second period, it would supply a convincing explanation of the change which then took place in the use to which the cobbled enclosure o ver n , ther it e s devotede yfac b f th wa obviouo e eo t n O s . swa reaso wha y previousld nwh ha t y bee nbarrack-yarda , calso y t li ma e w f ,i should have been all at once commandeered for storage accommodation. But, when the rampart foundation was increased on its inner side by, say fee9 l ,round al t , this would mean tha e freth te space, whics wa h already barely adequate for the needs of even a tiny garrison, would be reduced by close upon 8000 square feet. And an additional area of 800 e amoun0 th squar , be t ey s nearltha a fee ma , tis s t ywoula e db gained by appropriating the barrack-square and transforming it into the cobbled enclosure. That step, otherwise unprecedented and therefore difficult to account for, becomes immediately intelligible if it is looked upo s 'compensatioa n r disturbance.fo n ' Moreover, whil e prosth e 1 The abundance of stones may indicate either hasty construction or a scarcity of suitable sods regards A . secone sth d alternative perhaps i t i , t alwaysno s sufficiently realize heavw dho y was the demand which csespiticious construction involved. Mr Smith has calculated that as man acre8 s ya s must have been strippe orden di provido t r e origina th sod r sfo l ramparf o t Kough Castl eastn eo , west southd an , . * Tha originae th t l rampar beed ha tn similarly treated, whe s firse fornth wa t abandoned, is suggested by what the Report says as to the layering of the extension being "sometimes in continuation of the core " (supra, p. 263). ROMAN FORTS AT ROUGH CASTLE AND WESTERWOOD. 271 cowd an s were still being weighed, wha cannoI t t regarbu t conclusivs da e confirmatio s forthcominwa n e shap a mostth f n o ei g interestind an g quite unexpected discovery. The information given by fig. 4 as to the position of the north gate obviousls wa y unsatisfactory, inasmuc t indicatei s ha entrancn da e fully 50 feet wide. To clear matters up, we made search for the south kerb of the Vallum at the point where the gap is shown as beginning on the east. We struck it without difficulty and, following it along, found that, so far from ending where the plan of 1905 suggests, it continued for no less than 26 feet farther west. The break for the gate was thus 95£ feet distant from the inner kerb of the eastern rampart, while, as will be seen froe actuath m , "figl11 .openin s 20£gwa feet widefigure Th .s ei nearly twice as large as is usual for the corresponding gateway at the other Vallum forts, tha o gate s 1 th te must certainly have bee ndoubla e one.2 We noticed that the road passing through it had been repaired by relaying, exactly after the fashion with which the Military Way beside the cobbled enclosure had made us familiar. The double layer, however, measured only abou fee0 1 t t across e entrance widtth Th f .o h e had, therefore, been reduced by a half when the fort was reconstructed e seconfoth r d time. Incidentally e werw , e abl o accoune t e th r fo t erroneous impression conveyed by fig. 4, and to realize how fortunate we were in having chosen to work along the south side of the Vallum. On the north, which was preferred in 1903, the kerb, along with a good deal of the foundation behind it, had been deliberately torn out in post- Roman times for 26 feet. A a resuls f thio t s experience, it seemed desirabl o makt e e quite sure of the position and width of the south gate, and it was here that surprise awaited us. The gateway, as it is marked upon fig. 4 and as it appears to-day indee d occupatione ,ha th f do beed gatewae en n th .e th t ya But it had not been the original entrance. It had been formed by breaking throug e ramparthth s firsa , t erected d removinan , g 9£ feet of the stone foundation underneath. Its predecessor, the south gate of the earliest lay-out, had been 14 feet wide and had lain farther west, more nearly opposit e norteth h gate. Betweeentrancw ne d e nth ean e ramparth d beed ol ha tne th left intac fee3 r inches3 t fo t . Measured from centr o centret e e intervath , l separatin 5 feetg1 the.s mwa Owing, however, to the reduction in width from 14 to 9J feet, the e lateth edg f r o estree t fee7 withi 1 inche3 t s fore nth wa st farther away from the western rampart than was the edge of the earlier one. 1 At Bar Hill and Balmuildy it was 12 feet, at Mumrills 11, at Castlecary 10, and at Old Kilpatrick even less. Gadder wit fee5 h1 t approache t mossi t closely. 2 It would have been impossible to ascertain details without interfering too seriously with the heavy cobbling which extends all the way across. 272 PROCEEDING E SOCIETYTH F O S , MARC , 1933H13 .

;-. -^ » ______i L______-_-^(/^ -^v x El 1""" > ! " ""\^<%\\ v H = -I :8l'l'lH'Vl'1X o :::: O-- II F Pf^iig!< Oi 1 i: rr ^/|.ff^/Mffee I| jo j01H'l'1'VV&. .j}lMMMl!!!lillM'lM1ll!ri»;.:r-.'<' \? 4\ rr ^jiu? | .^'.''''''"''''Vr1 a. E- ^ < =r "^ ^ i O 0--

O--

°J- EQJ^ IO~" o UJ - _J< o -- -i\ i •: io"J o-- I oog 3| l^vmMMMMMMMMM.!.!.-,-,,, ^' 0< ;:| fSiiiiMiiii"inii.itu,,,.;,j,in,""",, /~^ o I s

sM ROMAN FORT T ROUGA S H CASTL WESTERWOODD EAN 3 27 . shalAe presently e sw orden i se l s gaio t r wa n t ,i thi s additional space that the change in the position of the gateway was made. As the two roadways, old and new, entered the fort, the contrast between them was striking. That upon the west had been admirably constructed and showed hardly any signs of usage. The gravelled and well-cambered surfac n havca e e born t littlbu e e traffic s beforwa t i e utilize foundatioa s da mixture th r f nfo eart eo sodd han o st thas wa t mak t ramparte pari parced th thao an t tf s o tl A upo . easte th n I , wrote in 1911: " The hard stones of the street that entered the Porta Decumana were worn into ruts. Manthemw sa f thos thes yo o a , yewh lay exposed during the excavations, must have been reminded of Pompeii." penninn I : g those sentences pern drawinow s - y wa m I , n go sonal recollectio d seef whao nha n I tsom e half-dozen years before. Last Januar s disconcertewa I y fino t d d that there wer stoneo n e s within the gateway at all, nothing indeed except a few shivers lying e naturaonth l surface hardln ca I y. think thamemory m t y was—or, rather, is—so seriously at fault as this condition of things seemed to imply. It is easier to suppose that the roadway was 'left open' in accordance with the disastrous policy I referred to at the outset, and that its stones have shared the fate that has certainly overtaken the interior buildings. With it all, however, the traces of the waggons have not been wholly obliterated. Even the natural surface shows two well-marked depressions, emphasizing the lines over which the wheels must have passed and repassed. The absence of wear and tear in the earlier gateway led us to examin e grounth e e wholn th froni d f eo t entrance more closely tha mighe nw t otherwis nd eo havha reaso d e ha distruso donent e W . t fig. 4 whe t i representn e roadwath s s solide a ywer W .e therefore hardly prepared to find that the two ditches had originally run past this fore th sid t f ewithouo t interruptio kindy an f . no Late , obviouslon r y when the position of the gateway was altered, both alike had been e piecea th f o s a ,l al filline feeta widt 6 s 2 r Th f .fillewa gfo ho n i d e th larg n o e t i boulderfee 9 f o e t showedi th f t t eitheo a s d Bu . en r wess virtuallwa t y untrodden e stoneth ;y loos la s e amon e soilgth . The 17 feet towards the east, on the other hand—that is, the part in latee fronth f r o gate—hat dvera y different stor tello y t s surfac it ; e was beate a firmnes o t n s which spoke plainl f mucyo h comind an g going. One could not help wondering why the western portion should have been filled in at all. Possibly it may have been because the old entranc lefs tewa open unti e alterationth l s insid fore eth t were com- pleted. 1 Roman Wall Scotlandn i (ed. 1911) 231. p , . 8 1 VOL. LXVII. 274 PROCEEDINGS OF THE SOCIETY, MARCH 13, 1933. Whether that was so or not, the new evidence has clearly an important chronological bearing. Before the gate was moved to the east, the southern entrance played a comparatively minor role as a channel of communication with the world outside. It is true that of the four nearese loop-line th th Militare s gateo th t t wa f t eo si o yn t WayBu . one who was making for the fort would use the loop-line at all, so long ordinare th s a y rout availables ewa ditchee Th . s coul bridgee db d with planks sortia contemplateds f i , ewa l otheal n i ;r circumstances wa t i s as well—or better—to leave them open. The situation was, however, profoundly modifie e transferencth y b de Militar th y f propeo e yWa r to an easier route (fig. 5). A new era was thereby inaugurated in the life of the south gate, which had henceforward to be used by supply- waggons t onl enterinn no , yo castellume gth t als e leavinn bu o o Th . git makeshif f plaiio t k bridge o longen s r sufficed. Something more per- manent and more substantial had to be provided, and accordingly a regular roadway was constructed. As this step coincided with the change covirsee e Militarth th f n i o y Way t musi , t have been taken no o reconstructiontw e e occasio th th e firs f th o tf o n s which have been so often mentioned, and it will be remembered that it was then that ditce th sout e fronn hi th f ho annexte gatth fillef s eo . ewa d in

Glancing back at the various alterations that have been described, we can now see that they were intimately connected. When the fort was re-occupied after its first abandonment, a desire to avoid a couple of very awkwar y bringindb t gradient e Militarme th gs y wa s Wa y alon ga lin e skirtin e southergth n sid -thf eo e enceinte. This involved ditcheo tw e sth immediatele fillinf o th n i g y opposit e soutth e h gateway. Simultaneously the gate itself was moved some distance towards e eastth . Her a edifferen t evebu tn more compelling- motivre s wa e commandantsponsiblew ne e Th . , perhaps becaus s disforce hi -eth t a e posal was too small for him to trust to the old-fashioned Roman method of challenging assailants in the open, determined to make the rampart top more effective as a defensive platform by increasing its width so as to allow greater freedom of movement. He therefore added an extension of 9 or 10 feet all the way round its inner side. One result of this, as we already know, was the appropriation of the barrack-yar r storagdfo e accommodation change e positioth Th . n ei n e soutoth f h gats anotherwa er informatio e arrangeOu .th o t s -a n interioe men th scanto f to o t s permiveri o ry t an y f positivo t e con- clusions beyond this general statement e statementh t Bu . t itself calls for some justification, even if it be of a hypothetical character. Whe studiee non plae s th considerd n (figan ) .11 larg e sth e proportion ROMAN FORTS AT ROUGH CASTLE AND WESTERWOOD. 275 e areoth f a withi e rampartth n s whic s occupie e centrawa hth y b dl grou f buildingso p e canno on t , wondebu t n werr me where e th e housed. In ordinary castella the long narrow blocks which served as barracks measured between 130 and 170 feet. But, when regard is had to e demandsth streetse th f o , t wil,i immediatele b l y apparent that nothing approaching that size could have been fitted in at Rough Castle t alla , excep angln a settiny b t ea whict gi h would have been hope- lessly out of alinement with everything else, and would besides have effectually prevente economican da beine us l g mad evef eo limitee nth d space that was available. At the little fort of Cappuck, however, Messrs Miller and Stevenson uncovered two barrack-buildings, each of which represente dtypa e conceivablthay ma t y have been adopter dfo Rough Castle.1 The larger, which it will be convenient to distinguish as Type a, measured about 96 feet by 17 over walls. The corresponding figures for the smaller, which I propose to designate Type fo, were 50 e followinTh . 22 g d calculation an thesw ho e o typet s a s s might have been employe t Rougda h Castle are f courseo , , subjecconditioe th o t t n that a minimum of 9 or 10 feet must be added to the end of each, to allow for the passage of the inter-vallum street. It is important, too, remembeo t r thae barrack-blockth t a castellum f o s werea rule s a , place principalis.da vi paralle e th o t l 2 If fig. 11 be consulted, it will be seen that the western edge of the via prcetoria, as it passed through the north gate, was about 108 feet distant fro e originamth lwese kerth tf bo rampart . Durin e eargth - three liesth f eo tperiods , therefore, ther havy ema e beeeast-ando ntw - west block f Typo thisn i se prcetentura,a th par f o t separated from one anothe streea y b r t occupyin same gth e positio thas na t whics hwa partly expose e n othe1903i dth rn . I portioe prcetenturath f o n e th space was less ample, the interval between the via prcetoria and the original kerb of the east rampart being only 95£ feet. Whether the first lay-out made any provision for barrack-blocks here is doubtful; havaree y th ama e held workshophavy ma e t accommodatei r o s e th d stores which were subsequently transferred to the cobbled enclosure. Wha s certai i tcontai d di s thatt i ni nf i ,barrack-blocks , these cannot have been as long as Type a. There would, of course, have been no difficulty abou e east-and-westh t t dimension f blocko s n f TypO o s . b e e otheth r hand, their breadth would have been such that there could t havno e been mor f ethe o thaeacn e mo non htransverse th sid f eo e street. That would have been wastefu e firsth t n i periodl t woulI . d

Proceedings, vol. xlvi. (1911-12), p. 450.

t HousesteadsA 1 2 , wher t righa t enclosure shape t se theth th angle e f , eyo ar it o st e made usuae th l arrangement impracticable. 276 PROCEEDING E SOCIETYTH F O S , MARC , 1933H13 . have been less so in the second and third, since the amount of ground left unused would then be reduced by the 9 or 10 feet that had been annexe rampare th r dfo t extension r eve o blocko e ntw f TypOn o s. eb may, therefore, have been inserted when the extension was made. Unless the garrison were cut down to an extent that seems unlikely in the case of so small a fort, something of the sort would undoubtedly be necessary to offset the reductions elsewhere. The two blocks of prcetoria,a vi e e westh r instance th f fo o Typto t ea , must have been shortened by 9 or 10 feet. This is admittedly all very speculative. We can see just a little more clearly e retentura.whe th e tur w no t n Her e distance th e th f o e original gateway fro e originamth l rampart-ker s practicallwa b e yth e west—10th n o e eas samth s 4a t n feeto e e ,th minuso t e lose th du s roundin e cornersth e samf o gTh .e typ f barrack-blockeo , something a little shorter than Type a, would therefore be as suitable for the one side as for the other. On the west, however, only a single block can have foun placa d e betwee intervallume nth e bace th streeth f ko d an t grou f stonpo e buildings which appea . elbowfigo n Theri rn 11 . s -ewa room in which to seek compensation when the rampart was extended. Rather than curtail the block still further, it was decided to move the gate eastwards, and at the sarae time to reduce the width of the entrance from 14 feet to 9 feet 6 inches, changes which meant a sorely needed gai f nearlno situatioyw 4ne | e feetnTh .thu s createease th t n o d coul e dealb d t with very simply. Barrack-blocks approximatino t g e questioth t knof o w no t wTypnho o d ou there a weree w W no e. man f theyo m ther havy ema e beed n ol while gate stils th eth n ewa i l position. Afte e reconstructionth r , however t wouli , possible b d o t e find comfortable accommodation for four blocks of Type b between e intervallumth principalis.a vi thio e d t i woulsth stree o T d d an t obviously be necessary to invade the latera prcetorii—that is, the space immediately alongsid e Headquarterth f o e s Building t i wil e d b l An . observed from fig tha 4 .e excavator th t f 190o s 3 faile discoveo t d y an r stone foundations there t probablI . riod di yt occu o thet r looo mt k for post-holes. This Nots alread muco t ha e n hyru greater length tha nI antici - t beei patedd n Ha briefer. I shoul, d have trie o brint d varioue gth s threads together by giving a detailed estimate of the extent to which r knowledgou f Rougeo h Castl s beeeha n advance e worth ky d b whos e results I have chronicled. As it is, I must content myself with saying that most of the riddles presented by the original Report have been satisfactorily answered I doub. t whethe e shalw r l ever learn very much more about the Antonine fort. Of the first-century occupation, ROMAN FORTS AT ROUGH CASTLE AND WESTERWOOD. 277 e otheoe significanc th ne th stil ar rf e o liliale handth w d f o an e, almost entirely ignorant. A word must be added as to the 'finds.' The lines on which we proceeded made it inevitable that these should be few in number. The most interesting was the Samian bowl which has already been described.1 There were a good many amphora fragments, including a handle — found near the south rampart of the annexe — bearin e letterth g s A-L-F-0 a stam, p which occurre t a Newstead d an d which is also known abroad.2 A small vase or jar (fig. 12), which came from a drain under the Antonine Vallum, in front of the fort, should also be mentioned. About half of it t togethecoulpu e db r fro piecee mth s that survived. It is made of dull reddish clay, but it has been covered

with a black slip, small patches of pig. 13. which still remain. The shape is suggestiv earle th f yeo second century data , e that woul consistene db t with the -'find spot,' for there is no reason why the jar should not be a 'survival.' Bound the shoulder are two girth-grooves, from the lowe f whico r h ther f decorationeo dependy wa y b , , line f doto s n i s engobe t iI s .moute 6|th inche outsidt n a ha s sha e hig d diametehan r of 4£ inches. Its most peculiar feature, however, is that in the bottom, a whicdiamete s ha h f 2o Jr inches, thera carefull s i e y made hole, approximately circular and f of an inch in diameter, like the hole in the bottom of a flower-pot. II. WESTERWOOD. The immediate issu t a Westerwooe s comparativelwa d y simple. Althoug e remainr leshth fa s e considerablar s e tha t Rougna h Castle, the position of the fort has been known from time immemorial. The track of the ditches on the south and west is still visible on the surface, •while eas th e lin tf th e eo ditche s betrayei s well-markea y db d sub- lone wald th gf en sidenco lrang e th f faren o ei m buildings. This being so I shoul, d hardly have felt justifie pushinn i d g enquiry furtherd ha , it not been for the conflicting accounts that have come down to us from eighteenth-century sources e varioun i Th .e son writert a e ar s 1 Supra, 258. p .

C.I.L., xiii., Pt. iii., Pasc. 1, No. 10002, 301 (Lezoux, Worms, and in the Museum at Mainz). 1 278 PROCEEDING E SOCIETYTH F O S , MARCH 12, 1933. holding tha greae th t t Ditch had usuals a , , forme dnorthere parth f o t n defence t ther curioua Bu s .ei s differenc f opinioo e n regardin greae gth t Rampart. On Roy's plan the east and west ramparts of the ' station' made ar abuo et t normae againsth n i t li t fashion. Gordo Horsleyd nan , on the other hand, show a gap at each of the two points of junction, thus leaving the northern front without any rampart at all. The description e alse e Militarcours t ar th varianceth a o y f f o o se Wa y . Gordon says: " Wha s veri t y peculia r e Causewahereth s i , y which goe sRamparts."e roun th t eas f no o o ydt s p i thi t To I s1e Fortth n o , attac y veran hy intelligible meanin s statementhi s thero i t g r eno , much enlightenmen t fro e accompanyingo m th e b o t t e g th plan t bu , story, , receiveconfuseis t i s sa d both suppor d illuminatioan t n from Maitland writeo wh , s thae statioth t t Westerwooa n s fortifie"i d d with a rampart and ditch, and the military way runs round the latter in a different position from what it does at most of the other forts."2 Horsley, again, while explicitly rejecting Gordon's view, affirms that the road "passe norte swalle th clos th y n hfort,"e y o , an eb sidth t f eo bu 3 traces he saw would seem to have been slight, for he does not mark them on his plan. Finally, Roy represents it as running right through the castellum between the east gate and the west, as it so often does elsewhere. witmord m ha h hi e d talthaa ha n k I I nyea a r ago Drysdaler ,M e th , proprietor, had emphatically endorsed Roy's version of the matter. He has farmed the land for more than thirty years and he annually makes acquaintance with the cobbling of the Military Way in the ploughing season. He had no doubt at all as to its entering the east gate and issuing from the west. But for the fact that it is always dangerous to ignore Horsley, I should have been disposed to accept this as conclusive. As it was, I felt that some investigation with the spade was desirable. Last autum fiele nwhicth n d i h 'for e mospotatoesn th i t f s lieo t swa d an , Mr Drysdale readily agreed to a little digging being done as soon as the crop was lifted. In September Mr Smith and I, along with Mr Mann, pai dpreliminara ysite discuso visith et o t plate scampaignf th no t I . was clear that it would not take long to discover the truth about the Military Way, and it occurred to us that with a little additional trouble we migh able b t ascertaio et exace nth t dimension forte th .f o s Work was begu n Novembei n d proceedean r d continuall r foufo y r weeks, despite the inclemency of the weather. High winds hampered the task of measurement, and violent rainstorms reduced the already heavy soil to a sodden condition. In the circumstances the plan (fig. 13) reflects great credit on all who assisted in its production. It forms the 1 Itin. Sept., 56. p . 2 Hist, of Scotland, . 175volp . .i . 3 Brit. Rom., . 170p . ROMAN FORTS AT ROUGH CASTLE AND WESTERWOOD. 279 basis of practically everything I have to say. The few scraps of pottery that were unearthed are too inconsiderable to furnish any information. And ther nothins ewa g else.

1^^^^^

~ VOTAR ___Y WA _Y

\^^5^^ ^;:'><><<>>>^^^ """"«.,„,nil,,,,,..' L«miumllilaA,.mW,iv1,»ia,l,uu»uw^

O O 5 10O IOO 2OOFEET I I I I I I I I I I I

Fore Westerwoodf Th tFigo . 13 . .

The north-east quarte castellume th f o r occupies i farmhouse th y db e with its garden and steading. Otherwise the site is unencumbered. Apart from some smal t significanbu l t irregularities, which wile b l touched on later, the fort had been of the normal shape. It had had e greath t Rampart fos northerit r n e Militardefenceth y d an Wa y; a principalis.vi e d serveth ha n s thes I a do respects tw e , therefore, 280 PROCEEDINGS OF THE SOCIETY, MARCH 13, 1933. the plans of Gordon and Horsley are misleading. Internally, the enclosure measured some 275 feet from east to west by 290 fro'm north southo t , givin n arega f nearlao acresy2 . As usual, ther d beeeha n four gates. Those on the north and on the south, although not exactly opposite to one another, had been approximately in the middle of the side whico t s h they respectively belonged e wess th t tha wa tn Bu o t. full fee0 y9 t farther fro soute mth h rampart than fro e Vallumth n mo the north. The eastern entrance was inaccessible. To judge, however, from the line taken by the Military Way, so far as it could be followed, difference th e therhavy ema been slightly greater t iI s. obvious that e Headquarterth s Buildin s faceha g d north s generalli s a , y (but tno invariably) the case in the forts on the Wall. The retentura, or space behind it, must therefore have been more than twice as large as. the prcetentura which lay in front. We may take it for granted that it had held the majority, if riot all, of the long barrack-blocks, in which garrisoe rane th fild f th eo kan n were quartered. The rampart of the fort was constructed in precisely the same way as greae th t Rampart turff itselfo s , restin;wa t thasubstratua i , n tis go m of laid stones. An examination of the junction of the two at the north- west corner brought out the fact that the foundation of the fort rampart rose 18 inches higher than the foundation on which it abutted, and overlappe abouy b t d i foo a t t (fig overlappine . 14)Th . interestn a s gha - ing implication. It almost certainly means that we have to do with the work of two different sets of hands, and that the legionaries who wer e builder e Valluth th f carried o smha s foundatiodit n pase th t contemplated site of the fort before their comrades—presumably the future garrison—had made much f anyi , , headway wit e particulahth r task that had been allotted to them. We know that elsewhere—at Balmuild d KilpatrickOl d an yr example fo , 1s —iotherwisewa t t Bu . confirmatory testimon whao t s ya t happene t Westerwooda furnishes di d by the Ditch, which runs straight on without a break, completely dis- regarding the road that was to issue from the north gate of the castellum. Moreover, there is no reduction in its width where it passes in front of e fort th approximatioo n , s dimensionit f o n e ditcheo thost sth f o e s surrounding the enclosure. The latter, as will be seen from the plan, were two in number for the greater part of the circuit. On the west side, however, a third was added between the Military Way and the Vallum. Whethe a similar r precautio e eass taketh twa nn sideo n , longeo n s i rt i possibl sayo et , sincI hav s e(a already indicated) that portion of the ground is beyond reach of the spade. It should be added soute thawestd th hn an to , particularl weste th Romane th n , yo s have

Proceedings,e Se vol. Ixvi.. (1931-32)f 7 22 . ,pp 1 ROMAN FORTS AT ROUGH CASTLE AND WESTERWOOB. 281 taken advantage of natural depressions in digging their ditches. And it will be noted that the double ditch was continuous in front of the south gate. Before seeking to account for this unusual feature, it will be well to give some description of the rest of the defences. e foundatioTh e greath f to n Vallum measure fee4 1 d t from kerb to kerb and was thus of normal width. It runs along a gentle slope, which becomes more sharply accentuated a little way in front. Con- sequently o providt , whicd t i witebe ha h shoult onca e eb d secure

Fig. Junctio14 . Wesf no t Rampar Fore th f t o twit Antonine hth e Vallum, showing mouth of drain from the Bath-house. d level e subsoian th , l beneat beed ha n t hi excavate a maximu o dt m depth of 2 feet. As an additional precaution, a single row of boulders had been laid along the unexcavated, and therefore higher, ground on the north at a distance of about a yard. The remains of the superstructure were represented by 18 inches or so of laminated soil. Some 16 feet west of the north gate a kiln had been constructed in its northern face. This was f course,o , post-Romand beeha n t i t bu , entirely built from remnant f Romao s n masonry (fig. 15). Thers wa e nothing aboumaro t s purpos t s datit i tk it r eo e very definitely. But, if one can trust the analogy of the similarly situated kiln at Mumrills, it must have been mediaeval. The motives prompting the choice of positio beed e samnha nth botn ei h cases d the e self-evidentan ,yar : 282 PROCEEDINGS OF THE SOCIETY, MARCH 13, 1933. e mounth f turfdo , lying read hando yt , offere a conveniend t backing, and the existence of the hollow of the Ditch would go far to ensure a steady draught. The foundation of the fort rampart had been laid upon the natural d thusurfaceha s d beean , n much more extensively damagee th y b d plough. So far as could be jiidged, however, the method of construction had been otherwise identical. Both kerbs had usually survived,

Fig . Post-Roma15 . n Kiln, built e intAntonine bodth o th f yo e Vallum.

but sometimes the stones that had presumably lain between them had been torn out. From the north-west corner to the south gate, and dowy e easwa muco ns th e tf agaihth o sid s l coula en al e opene b d d up, the stone base had been approximately 16 feet wide. Between the sout e south-weshth gatd ean t cornebeed therd ha n An t ei ronl . y14 was another peculiarity here. Simultaneously with the reduction in e rampare widtth th f ho t there bega nnarrowina berme th f go , which ultimately shrank from 9 feet to about 4^. The south-west corner was also curious e pla th a loo n t s a ka wil, l show. Althoug e outsidth h e appeared to have been rounded, the inside had been almost square. Of that there could be no manner of doubt, the inner kerb being singularly well defined (fig. 16). The effect of this arrangement was to ROMAN FORT T ROUGA S H CASTL WESTERWOODD AN E . 28S

push the outer edge of the rampart so far forward that the space separating it from the lip of the ditch outside can hardly have been more than a couple of feet. Except at Rough Castle, where the extension of the rampart made e conditionth s exceptional I canno, t recollect anywhere eth elsn o e Antonine Wal n exampla l a ber f exiguouo o mes s 4y a £s sa feet o t ,

Fig. 16. Inner Kerb of the Rampart at the north-west corner of the Fort. thad nothinAn t thi. 2 sf go lac f conformitko o t e hardln du ca y e yb mere caprice seems more than ever e certainremembereb t i f i , d that e 4£th feet ber s massociatei d wit e rampara hsecto th f o r t whose breadth is 2 feet less than the average for the rest. A moment's study e plaoth fn ) will(fig13 I .think, e wha ,se e actua enablth to t s lu e motive was. The fort, as outlined by the rampart, obviously sits rather uneasily within the framework of the inner ditch, particularly at the souther absencn endThe . propeeof r co-ordinatio unmistakablso nis e tha I havt hesitatioo n e acceptinn ni e suggestiogth y b n e madm o et . CuriO . MeA r whe broughe r Smith'nh M e t m ts no drawing s wa t I . 284 PROCEEDINGS OF THE SOCIETY, MARCH 13, 1933. a clear field that awaite soldiere d th erecte o wh se Antonin dth e fort, when they too p theiu k r task. What they mad e inneth e f theio r r two ditches was already there, and they had to adapt the line of their rampart to the limitations which its presence imposed. A very instructive parallefoune b t Cappuck.n a d ca l 1 Are we, then, justified in adding Westerwood to the lengthening lisf Valluo t m site whicn o s h tracesn Agricolaa f o n presidium have com o lightt e givo T emphatin ?ea c answe e affirmativth n i r e would perhaps be unwise. But the prima facie probability is undoubtedly strong. And it may well be that the suggestion furnishes a clue to the anomalous behaviour of the two ditches over against the south gate. If the inner one has belonged to an Agricolan fort, it is easy to under- t i shoul stany dwh d ignor n Antonina e e entrance; andt i wer f i ,e already continuous o goon , d purpose woul e serveb d y makinb d a g thae s addeon wa t e dbrea th outside n i ke argumen Th . s noti t f o , course, conclusive, as the behaviour of the ditches at the south gate of Rough Castle wonderfull a shows. t Bu y close analog citee b dn yca

from Bar Hill. There th2 e two ditches on the west side run past the f gatetheri s a e, were questio gatth o n en i e t allna excavatiod an , n s proveha dd originall thaha t thia o t stw y poine e inneth th tf o r formed part' of the Agricolan lines. At Westerwood it is, of course, impossible even to guess at the shape of the earlier enclosure, if earlier enclosure there was e mus W conten.e b t t with having obtained pre- sumptive s formeprooit f o f r existence. interioe Antonine th Ath o st f o r e fort ther saide e littls ei b W .o e t have seen that most, if not all, of the barrack-blocks must have lain to the south of the Military Way, and we know that part of the Headquarters Building must be buried beneath the south-west corner of the farm-steading. It may be taken for granted that there was at leasgranare on t r storehousyo e alongsid e Principiat th no f s i o e t I . e eastunlikel r twenty-fivth fo ,n o s y tols thawa dewa t i tI year o sag that a quantity of blackened wheat had been discovered there when the foundations of the farm-buildings were being dug. Lastly, the recent excavations showed that there had been substantial buildings of stone within the north-west corner of the castellum. There was no time to follow up a fragment of walling (fig. 17), which ran north and south within a foot or two of the west rampart. But at the same distance behind the Antonine Vallum the spade brought to light 90 a e lonsid f on eo gfee f narroo t w building. Thiextended ha s d westwards fro e nortmth h gate, clos whico t e e retur th ha pars f o wa nt 1 Proceedings, vol. xlvi. (1911-12), p. 450. 1 Supra, 273. p . ROMAN FORTS AT ROUGH CASTLE AND WESTERWOOD. 285 also uncovered. The wall was well constructed, and varied in thickness from 2 feet 6 inches to 2 feet 9. That it represents the remains e Bathoth f s certaini s . From that poin f vie o e te shap th wth f o e building is significant, and not less so is the drain which passed under the rampart into the innermost ditch directly opposite its western end (fig. 14) comparisoplano A . tw se wilth f lno show tha Bathe th t s occupied exactl same yth e latte th positior Hillt r A Ba for . t na a t

. WalFig17 f Buildin. o l g withi westere nth ne Prcetentura parth f o t f o the Fort. space of 20 feet had been left between the west wall of the Baths and the west rampart, doubtless to give room for a corner-tower. If there was a somewhat similar arrangement at Westerwood,1 the bath-building there woul aboue db fee0 t12 tr Hilonl d lon Ba an s lagainsy gt a a 0 15 t 8t Croy0a . These difference sizn e roughli s ear y proportionate th o et difference e thre e areth th f e ao n fortsi s thereford an , e presumablo yt strengte th theif ho r garrisons. The emergenc f suceo hlona g stretc f wallino h n thigi s particular position, and only a short distance beneath the surface, brings us back e starting-poin th o what s r wa wholt ou f eo t investigationt no t i s I .

1 The question could not be definitely determined owing to the presence of a mass of tumbled stones which it seemed best not to disturb. 6 28 PROCEEDING E SOCIETYTH F O S , MARC , 198313 H .

at least possible that Horsley's view as to the course of the Military Way may rest upon a mistaken interpretation of the line of stones, which may well have been visible, cropping up through the turf, when he visited the fort two centuries ago? He was not in the habit of committin ge surb himseld somey ha tha ema e fh e t rashlyw d an , ground r assigninfo s o vers g y unusua e Romaa routlth o t en road. Nor is it at all unlikely that Gordon and Maitland had still stronger justification for the entirely different account which they give of the matter. What attracted their attention was probably one of the loops or by-passes of which I have spoken in connection with Rough Castle. t WesterwooA do unnecessar n ther s wa e y e saveclimb o dt b from. The only thing to be avoided was the actual passage through the castellum. In the circumstances the loop would be short, beginning just befor s reachede forth ewa t d hugginan , e outeth g r ditch until the main Way was rejoined a little beyond the west gate. This is not mere guess-work r SmitM . h tellthae m sn carryini t ga trenc t ou h fro south-ease mth t corner orden i , o satisft r y himself that therd eha only beeditcheso ntw f stonesfounde a thiro h , t d dno be , , ditcha t bu , closely resembling the bottoming of a Roman road. III. A POSTSCRIPT. Some time ago Mr John Clarke, in the course of a conversation about the work on which he was then engaged at Gadder, suggested to me tha e evidencth t e fro mReporte th Roug n i t , hpointeou Castlet se ds a , to there having been a raised walk on the inner side of the rampart there arrangemene objece th Th f .o t t woul enablo t e ddefendere b eth s o exercist a certaie n degre f commano e d ovee grounth r n fronti d , without exposing themselves undul e missileth o yt f assailants o s e Th . suggestion was naturally present in my mind when I came to examine e inneth r 'margin e stonth f eo ' foundation s consideratioit d an , n a e revieknow th o ultimatel t f wo e n m fact d le ys relatine th o t g rampart e nine wholth th ef f o so efort s that have been mor r leso e s thoroughly excavated along the line of the Antonine Wall. In doing so I inevitably wandered still farther afield, so that some of the conclusion I shav e reache e perhapar d s capabl a mor f o e general application t mayI . , therefore usefue b , summarizo t l e them. Nothing is more striking about the defences of the forts on the Wall than e nintheith ef r o case variety x e ramparf si th so n s I . wa t f clayo s . wa Althoug t i e s ol on f al hn wa stonei t i d o an , turftw n i , were built at one and the same time and were integral parts of one e samth ed schemean , ther s evidentlewa fixeo yn d typ whico et h their designers were expecte o conformt d . Castlecary r instancefo , s wa , ROMAN FORTS AT ROUGH CASTLE AND WESTERWOOD. 287 give nwala f stoneo l , clearly because there happene conveniena e b o dt t outcrop of suitable rock in the immediate neighbourhood. Clay, on the other hand, was preferred at Mumrills, where sufficient supplies of stone or turf were not easy to come by. And, even when the material employed has been identical, there are differences of detail—some of them by no means unimportant—which prove that individual com- manders were allowed a surprisingly free hand. The point needs emphasizin e presenth t a g t juncture. Despit r JameM e s Curie's wise words of caution, written more than twenty years ago,1 there has recently bee ntendenca classifo yt remaine th y f Romaso n a fort n o s more or less rigid chronological basis. When I came to close quarters with Mr Clarke's suggestion, it seemed to me that this would not be a very hopeful method of approach, and I deliberately selected another. castellume Th r permaneno t fort was f course,o direce th , t descendant e temporaroth f y camp begio T . n s girt withwa , t liki , s prototypeeit , wit a hrampar f earth o e tprimar Th . y purpos f thio e s particular element of the defence was to bar the way against assailants. But it was not designed for passive resistance merely. It had to provide the defenders wit hvantage-grouna d from which they could retaliate upon therefors attackerse wa th t i d e an vita, shoulp l to tha roome e db th t y enoug s o admibeinht it f go t effectively manne n e evena th f o n ti d assault. Gellygaer, which was probably erected towards the end of the first century, offers a most instructive example. The rampart there was almost 20 feet wide at the bottom and, as there "was a retaining wall on both sides, it is hardly likely to have been much narrower at the top. Such a double revetment is of rare occurrence. More usually

the inner side is marked by a kerb, which has supporte2 d the foot of a sloping bank and which indicates a width of between 30 and 40 feet e ban e baseth th kf feeI 5 t .were1 a t r fee2 o 1 ,2 t r 1 sayhig o d 0 h1 ,an broad at the top, the slope would rise at an angle which would make it unsuitable for a ramp. Ascensus valli, such as Hyginus postulates for the temporary camp, would therefore be as indispensable as they woul e whe b de inneth n r s verticalfac wa ee revetmen Th . n fronti t , t i shoul e explainedb d , would ris a elittly abov e wa rampare th e t behind, thus actin a breastwor s ga r parapeto k . Originallf o s wa t yi e firsth tf o centur-wood d t beforen bu , ye th ewoo s beindwa g replaced by stone. I can remember no more interesting or more convincing illustratio transitioe th f no n tha characteristicalle nth y lucid description 1RomanA Frontier . alsCf o . Haverfleld'f 8 Post,2 . pp s not Ward'n i e s Roman Fortf o Gellygaer, p. 38. 2 Ribterling, however, believe sbees thaha n t i tmuc h more common than might appeare Se . O.R.L., , Kastelvol3 . ii . l Wiesbade t CaerleonA (Nr . f originae 8 . n1 th 31) . ,pp l rampart, which claf dated o ys an swa from circa doubla d A.Dha e, revetmen75 . timberf o t . 288 PROCEEDING E SOCIETYTH F O S , MARC , 1933H13 .

which Fabriciu se evolutio th give f o sf 'Mauero n d ' Wall'an t ' a Urspring.1 Scotlan n sho dca t w leasa example on n earthet a f o e n fort which s demonstrablha y been reveted wit significans i h t woodi d an , t that i t should be a fort which was in all probability erected for the first time e reiginth f Antoninuno s Pius, hal centura f y after Gellygaer. When Birrens was excavated in 1895, interest was in the main (and very rightly) directed to the interior, the recovery of the plan of which was destined to be a landmark in the progress of Romano-British studies. The rampart was completely overshadowed. Nine cuts were indeed made through it—not all of them, unfortunately, carried down to the original ground-level—and profiles of these were drawn by Mr Barbour. They are described in the text of the Report, where also the drawings are reproduced. But they were never critically studied, nor was any serious attempt ever made to grasp their implications. Nevertheless, even as they stand, they tell us quite enough to make it fairly certain that an ad hoc investigation—which could still be carried out, and whic a costlh e woulb yt undertaking—wouldno d yield much valuable information. Thus, there woul difficulto little n db r eo findinn i y t gou whether there were stone towers at intervals, as one of the sections would seem to suggest, and whether these were connected by a terraced •walk, represente e fragmentth y db f "o s polygona l pavements," which appeare separato tw t da e points.2 For the present we are concerned only with the general shapfe of e ramparth witd s woodean thit n revetment e sectioTh n whic. o n h e evidence latteth th r r fo erests—evidenc e whic s hitherthha o passed unnoticed—wa e mosth s t complet e ninth ef o e(fig e centrth . 18)n I e. stona s ewa bottoming fee8 1 , t wide, above whic e ditc spoie th hth f h o l had been heaped. At a distance of 11 feet behind was a well-laid stone kerb. In the Report the kerb is assumed to be the outer edge of a flat terrace alonn , whic ru innee d gth hha ramparte r th sid f eo . Obvi- ously, howeverslopine e 'heelth th f s i o 'g t i ,bank , whic shoule w h d naturally expect to have terminated here. Its face was cleared for 45 feet s fas E(figa a r,) " .a littl18 e square recess sidese th , , backd an , botto f whicmo h wer f singleo e flat slabs, each abou fee2 t t square. . . . e arm r Th sideo s s were suspensioth e r notchefo f f i vesselso n s da ; charcoal was found in the bottom and near it."3 Mr Barbour's enlarged sketch (fig. 19) enables us to visualize the recess distinctly, and we need hardly hesitate as to its meaning. The notches were not "for

O.R.L., vol. vi., Kastell Urspring (Nr. .f 66a)5 . pp ,

1 See Proceedings, vol. xxx. (1895-96), pi. ii., Sections O.P. and Q.E.

32 Ibid., . 9. 8f pp ROMAN FORTS AT ROUGH CASTLE AND WESTERWOOD. 289 e suspensioth e suppore e lowesth th f th vesselsr no f f fo o o tt bu " wooden steps—perhaps a movable ladder—which had given access to e rampart-topth recese Th . s wase th fact n i ,e f beginnin o th , e on f go

STONE BOTTOMING KERB SECTION OF RAMPART

^! £

•S\) JUrVT\T\T« r\'*«*,rv t«».*XTs *«\<\TStSTrVT ^ Jv-f^*•-•"•• i ^ AVv^rvf^ ELEVATION OF KERB IO O 10 2O 3O 40 FEET

Eampare Th . Fig 18 f Birrens.o t . ascensus valli, require e steepnesth e slopiny th b d f o sg bank. The " charcoal" we may suppose to have been merely decayed wood. Thu se inne mucth r e character fo hth e cluface e o th Th t e . f o r outer face is furnished by the 'slot' that is visible in fig. 18, about 6| feet in front of the bottoming of stones. It rn is 2 feet deep and rather more than 1| wide ,; INI at the top. That it has been a post-hole and / ——* — '— ^ ) i t palisadpara f no o t e trenc proveds hi t no , merely by its shape and size, but also by .//. e t faccontinuousth no t s thai t i t : onln yi >/ ... JKH (W/MMfflffiffi' //' e ninothee th on ef o rsection s anythinswa g PLAN OF L SECTION OF £ 1 kine oth f d observed.safely ma ye infeW r 0 14 81 2 3 4 FEET that, exactly as at Urspring, the revetment .mil 1 1 1 had consisted of horizontal planking, nailed Fig Reces. Rampare 19 . th n si t seriea o t f uprightso r valli.cannoe o s W t at Birrens. say positively how high it was. But, in order to provide a serviceable breast-work t musi , t have fee 4 rise r to abovn3 rampare levee th eth f lo t top. This, again hardln ca , y have been mor fee2 1 e t r thao abov 0 n1 e e bermy excesth An .s over that would have lessene s valuit dr fo e offensive purposes by increasing the extent of the ' dead' ground, within the limits of which the assailants could shelter themselves from missiles.2 Of the width of the fighting platform we can say nothing, sectioe Th 1 thas ni t lettere dplate Q.Bth en o .alread y referre. dto 2 See von Groller in ROm. Limes in Oesterreich, Heft ii. 25, where 3'00-3'20 metres is suggested as a maximum height for the wall of the fortress at Carnuntum, exclusive of the battlements. Brecoe th t A n Gae revettine rth g wall still stand fee1 s1 t high. VOL. LXVII. 19 290 PROCEEDING E SOCIETYTH F O S , MARC , 1933H13 . except that 15 or 20 feet is probably the utmost that need be allowed for. On the other hand, the width of the rampart at the base can be calculate r mordfa e accurately thadons n1895wa n ei . Measured from e outsidfeete post-hol8 th 3 th e s ,outsidf kerbe eo th wa th o t et f i , o e or precisel clae e sambase th yth y s f thath emouna eo f o t d behine dth wall at Newstead. So far as I am aware, there is no fort in Scotland of which we can say with confidence that it has been an earthen fort with a stone revet- mentScottise Th . h forts thae sometimear t describeo s s d really fall into quit a edifferen t category t ther e Bu numerouar .e s examples e Bordesoutth abroadd f o h an rcharacteristice Th . trua f eo s revetting wals e relativelthai ar lt i t y narrow—seldom more tha fee4 n t thick—and s lefi tha t i t roug d unfinishean h d e inneoith l r face. e Brecoth t A n Gaer r instancefo , e averagth , e thicknes fee3 5 s t wa s e sidinchesth e d nex rampare an , th t s 'tusked,o ensurt wa tf i a es a ' firme re mas th gri f supportedearte f o sr b po fa o h t s i thas t I wa .t from unlikely that the uppermost part of the revetting wall—that is

the portion of it which served as a parapet—was narrower than the( lower part against which the weight of the rampart rested, for a parapet of 4 feet broad would have been ill-adapted for the use of missiles havy l thiAl ma se. helpe hasteo t d e advennexe nth th tf o t e evolutiostagth n i e f fort-defenceso n e timth ey B tha. t Lollius Urbicus entered Scotland the tendency to break up the 'composite' rampart into its two constituent parts was fully developed. We may trac e fortuneeth f eaco s f theho m separately. As regards the rampart proper we need not travel beyond the Antonine Wall for illustrations. By employing a different material s transformewa t i d int n independena o t structure t MumrillsA . r fo , instance, it was a mound largely of clay, resting upon a carefully laid stone foundation, 12J feet broad. More often it was of turf. The co- herent quality of sods was sometimes turned to account as a means of holding masses of loose earth in place; cases in point are its use as ' Vallumkerbin e th Hadrian'n n o i g' cradlina s a Wal d gan l e foth r original earth rampart at Urspring. It was but a small step from that to building it up as a regular wall. Occasionally, as in the murus ccespiticius at Appletree, the sods were simply laid upon the natural surface a rule s A , . however , a stonther s ewa e foundatione th n i s a , Antonin ee fort th Wall sn I .wit he immediatelwhicar e w h y concerned e breadt th e foundatio, th f ho n varied, ranging fro fee2 t m1 a t fee0 r 2 Hil t o Ba t l(an d ultimatel t Rouga ) y38 h Castle e superTh . - structure, whethe clarof turfyor , must have feebee10 tleasnat or 9 t high if it was to serve its primary purpose of keeping assailants at ROMAN FORTS AT ROUGH CASTLE AND WESTERWOOD. 291 bay. This would allo wa minimu p e mflato th t widt feer 6 fo f to h which served as a fighting platform. It is tolerably certain that there would be a parapet, either of timber or of wattles. The conversion of the retaining wall into an independent barrier was an even simpler and more natural process. All that was needed was to double its width and impart a finish to its inner face. Allowing a rampart-walke th r fo p minimuto buildere e th ,th fee 6 n f mo t so would hav e sparo battlementea t roo r d efo m an d parapet. e fullThath -s i t grown ston e secone th for f o dt centur t Castlecara fine t w i d s ya d yan Balmuild t Newsteada d yan . But, even whe stone nth e attaineford ha t d s majorityit , ther s verewa y ofte mouna n f clad o r eart yo h behin. dit It is customary to speak of this as if it were still the rampart, with the unfortunate result that the independent stone wall is apt to be treated t weri f i emer a s a e revetment equalls i t I . y misleadin reverso gt e eth position and to look upon the mound as designed to support the stone wall well-builA . t wal f stoneo l fee8 , tfee 2 1 broa t d highdan , needed no backing of earth to ensure its stability. How, then, is the presence e moune explainedoth b f o t d f I tha? t questioe answereb n ca n d satisfactorily shala gooe w e , db l deal neare a righr t understanding hitherto tw r o oe puzzlinoon f Vallun g featureow r m ou forts f so . Where a pre-existing earthen fort was rebuilt in masonry—by which I mean something more tha e substitutioth n a ston f o ne revetment f timber—io e foon r t woula perfectl e b d y intelligible proceedino t g rampard leavol e eth t standin walw s frontn i i ne lerecd t e I gan . th t less easy to see why an earthen mound should have been piled behind the wall of a stone fort that was built on a virgin site. Yet there is one very obvious reason. It offered a convenient means of disposing of the upcast from the ditch or ditches, a problem that must have been more serious tha alwaynis s realized. Ritterling calculates tha Wiesbadentat , where there were two ditches, each linear metre of the enceinte meant about 16 cubic metres of upcast. And the Romans would not have

been the practical people that the1 y were, if they had been content to a mer e b dump.' e t i t 'le Ther gooe ar e d ground r believinfo s g that they turned it to excellent purpose. On that head the description which von Grolle e fortressere wallth giveth f f o so st Carnuntua s d man Lauriacum is most illuminating. With the aid of fig. 20, which is based upon one of his diagrams, I will endeavour to summarize the salient point arguments hi f so . Fundamenta clear-cua s i t i o tt l distinction betwee 'Wallgang'e nth or rampart-walk and what he calls the 'Wehrgang,' a word which I 2

O.B.L., Kastel, 3 . ii l Wiesbaden (Nr. 31)18 . ,p

1 2 Bom. Limes Oesterreich,n i . f 1 Hef3 . ii t 292 PROCEEDING E SOCIETYTH F O S , MARC , 1933H13 . propose to translate as ' sentry-path,' although the connotation of the English expressio narroo to s ni permiwo t s bein it f go t acceptea s da wholly satisfactory rendering formee p s thaTh to . wa e rt th par f o t of the stone wall which lay immediately behind the battlemented parapet whicd servee fightinw an , th h no s da g platform from which e activth e defenc th par p conducteds f to o t wa ee th lattee s Th .wa r of the earthen mound, which was some 3 feet lower than the wall. Soldiers standing r movinupoo , it n g alon , woulgit muce b d h more effectually sheltered from snipers than if they were on the rampart-walk, on to which they could be ready to mount if a direct attack were launched. Meanwhile they would hav a egoo d l thas vieal wa f tw o going on beyond the inner ditch and might even have an opportunity

NVV.\.>'VV.N' \XXX\\\\\V\\\\V

Fig. 20. of hurlin gmissila e home earthee Th . n mound, then, fulfille definita d e and.a very useful function. Nor is this mere theory. It is borne out by positive evidence. Except in so far as the dotted lines are concerned, the diagram reproduced in fig. 20 is a representation of what von droller actually discovered at Lauriacum.1 The earthen moun s belod i presen e levee th w th f o l ts surfacwa d ean not visible until it was exposed by the spade. Knowing by his experience t Carnuntua m wha expecto t t basind an , s calculationghi ascere th n -so tained position of the berrn, the excavator made up his mind beforehand as to the amount of soil that would have to be removed before he reached the ' sentry-path.' That his estimate should have turned out correce e b sures th o t s i tte soundnes prooth s f hypothesiso hi f f o s . Incidentally remarke ,h s tha angle th t t whicea moune hth d leavee sth edge of the intervallum street produces a slope which is gentle enough for an easily ascended ramp. A reference which he makes to the

. MilleN . S r r mad cit.,. M . 1Op verea 5 Hef . y xi tsimila r discover t Yorkya . There, however, the wall may have risen from 4 to 6 feet above the level of the ' sentry-path.' See Journ. of Roman Studies, vol. xv. pp. 177 f. ROMAN FORTS AT ROUGH CASTLE AND WESTERWOOD. 293 sculptures on the Column of Trajan is very relevant. Several of the scenes depicted there might be cited in confirmation of his view. In particular there is one in which Roman auxiliaries are seen defending a castellum agains n assault a ta hord y f b Daciano e s (fige . 21).Th 1 front wal s battlementedi l e ar t i d n ther e an soldier,ma th e o wh s evidently on the rampart-walk, for they are visible from a little above

. RomaFig21 . n Auxiliaries defendin forga t agains Daciaa t n attack.

the knee. Those who are guarding the side wall—only this and the front wall are within the range of vision—must be at a considerably lower level. Despit absence parapety th e an f eo moro n , e than their head shoulderd s an seen e b n . sca My knowledge of the stone forts south of the Tweed is not nearly detailed enough to justify me in generalizing about them. But I may be allowe refeo dt Housesteadso t r . stona Thas fule ewa t th forl n i t sense of the term. Its walls were as carefully finished inside as outside.

Cichorius. 1C Reliefse Di , der Trajanssaule, Taf. xxiv. (Bild xxxii.) sculpturea n I . d relief o f this kind it would be idle to look for precision of architectural detail. The essential point is the difference in the height to which the bodies of the two groups of defenders are visible. 294 PROCEEDING E SOCIETYTH F O S , MARC , 1933H13 . Whe openes fore ncenturth a wa t p du y ago ,bana earthf kf earto o r o ,h d clays clearean wa , d away from behind such part f theo s s mwera e excavated. It seems to have been 12 or 15 feet wide and to have been supported by a retaining wall. Mr- Bosanquet, in his admirable account of his own diggings at Housesteads, writes: "Hodgson observed 'a terrace, mad f eart eo clayd han , whic fron hra m towe turreo t r t along the inside of the wall to the height of about five feet above its foundation,' and noted that the insides of the towers of the gates and of the turrets betwee e cornern e wallth the th d f msso an were fille witp du h clay to the same level."1 Does not this seem to be a singularly close parallel to the ' sentry-paths' at Carnuntum and Lauriacum ? Thu s retury e castellae Antoninarmedth th ma o n e t o w , e Wall. They fall into two groups, and I will deal first with that which consists o stontw ee oforts—Castlecarth f d Balmuildyan y e . ar Thos o wh e familiar with what was found at Castlecary will remember the kerb whic n alon ra he norte insid th gth f heo frondistanca t a feet6 f teo behin e wallexcavatore th dTh . s were avowedlI d yan , baffleit y b d some timabandoneo eag dsuggestio a t forwar pu d 191n o di ha t ns I 1 a its possible significance.inclinew no o thinm dt a I k 2 tha t realli t y indicate sentry-path.' sa face tTh tha' t ther apparentls ewa y nothing of the kind on the other three sides is hardly a serious objection, since it was from the north that danger chiefly threatened. Nor does its seeming narrowness, as compared with what was found at House- steads, present any real difficulty. If the 'path' were of sods with a ' duck-board, very yma s wella ' have bee e case th ne sid ,th e nexe th t interior would be practically vertical, so that a space of 6 feet would be ample. At Balmuild a 'sentry-pathe casr yth fo e eves i ' n stronger. Such evidenc s surviveea d wen shoo t t w t Castlecarythata s a , wale d th , ha l been carefully finishe innee th n r do side . This, combined wit s thickhit - ness, place castellumthe scategor the in stonof y e forts properlyso r MilleM called t r Bu note. d certain features which might easily have escaped the eye of a less acute observer and which seemed to him to point o thert e having bee a nbank , averagin fee0 2 gn breadthi t , behind. Accepting the current view,3 he took it for granted that it had been a backin f earthgo , intende o support d e masonrye kinth tth dy B . permissio Glasgothe of n w Archaeologica MilleMr l Societof r and y himsel e diagrath f m embodyin interpretatios ghi reproduces ni d here as fig. 22. Such an arrangement as it indicates would perhaps be possible in a stone fort whose wall has been built immediately outside

1 Arch. Ael., N.S., vol. xxv. p. 246. » Roman Wall in Scotland (ed. 1911), p. 209. E.g. Jacobi s BomerkastellDa , Saalburg, , figx. .d iii Tafan . . ix . 3 ROMAN FORT T ROUGA S H CASTL WESTERWOODD AN E 5 29 . of an already existing earthen rampart. For a stone fort built on a virgin site it seems to me hardly conceivable. That it was not con- tinuou a stron s i s g argument agains s beinit t o a gbackingn s i t I . argumen l againsal t a s beint it t g'sentry-path.a bane f eartTh ko ' h and clay at Housesteads was interrupted by the latrines. At Balmuildy a bath-house blocked the way on the west side, to say nothing of an e south.th ove n o n1 Thae 'path e basth th tf eo ' should have bee0 n2 feet broad shows tha t musi t t have bee nmouna eartf f eartdo o r ho h clayd an .

5*51 Fig. 22.

The other seve treatee nb fortsingla y s s da ma sa e r groupfa o S . e presencth r absenceo 'sentry-patha f eo s concernedi ' e differencth , e betwee a ntur f superstructur f clao s immaterialyi e on d an e . What matter e breadt e th ston th s si f eho foundation Miller'r M . s accounf o t the excavations at Old Kilpatrick gives no hint of anything abnormal being observed there, and I can vouch for there having been nothing sore ot th Mumrillsfa t , Westerwood, Cro r y Hillt CroHillBa A .r yo , r HilBa l t thera Hild e an lwer e numerous hearth se ker closth bo t e e rampar oe th westf th n , o twhil t Westerwooa e r Hile Ba th l d dan long Bath-buildings were within 2 or 3 feet of it on the north. If definitely negative evidenc s lackinewa t Mumrillsga e searcth , h there 1 Mr Miller suggests (Balmuildy, p. 11, footnote 2) that the bath-house may have had a flat timber t rooftherBu evidenc.s e i shoo et w tha roofe th tbath-housef so s were generally arched, and, where the risk of fire was so great, it is difficult to believe that they were ever made of timber. 296 . PROCEEDINGS OF THE SOCIETY, MARCH 13, 1933. was thorough enough to make it certain that there was no blank space whic sentry-path' ha ' could have occupied. There remain Rough Castle and Gadder I hav. e already dealt wit e formerhth inferrinn I . g that wito d h o t a reinforcemen ther s d ha whawa e e e ramparw tth f o t t itself, I was guided not merely by the appearances described in the Note but also by a more general consideration. With a rampart that wa fee0 s2 tbase ' sentry-pathwidth e a t ea ' would have been useless. To obtain any view of what lay on the other side, it would be necess'ary to clamber on to the rampart itself. A 'sentry-path' from which it was impossibl evee e outese nth o et r ditch would have been something like a contradictio n terms i nr tha Fo t. reaso I nbeliev e tha a similat r explanation holds good at Gadder, where Mr Clarke has discovered that ot leasna t three sides cleathera seemingld s ran e wa y unoccupied space fee0 o1 ft behin e rampart e kerth s truth df i bo t eI .tha t there ear no signs of a stone foundation for the sods of an extension to have rested upon. But the surface is hard and they may have been laid on t directlyi , jus s thea t y wer t Appletreea Antonine fronn th i f d o t ean e Vallu t mRouga h Castle r sodd stoneo ; y havan s ma se been inter- mingle s theda y apparently wer Rouge somn ei th f eo h Castle sections.