THE ROMAN CAMPS AT AND GLENMAILEN. 317

V. E ROMATH N CAMP T RAEDYKESA GLENMAILEND SAN . BY GEORGE MACDONALD, C.B., P.B.A., LL.D., F.S.A. SCOT.

befory abls la Societe o th et wa A I yeaaccounn o ya ag r f certaio t n excavations which a Research Grant from the Carnegie Trust for the Universitie f Scotlano s enabled mako dha t e edm alone th e lin f gth eo Roman Wall from Forth to Clyde.1 In normal circumstances this investigatio nhavw woulno e e beeer d n completed. But, since th e momentous autum f 1914no , progres f necessito s ha s y tende becomo dt e increasingly slow; in the present national emergency it seems wrong to evek as nsingla e workma devoto nt energies ehi unproductivo st e labour, except on the rare occasions when seasonal conditions preclude the possibility of more useful employment. It is true that the thread has t bee e contraryno nth lostn O .a goo, d dea f interestino l g materias i l gradually being accumulated meantimee th n I . , however e furtheth , r repor positioa n ti whichope d e brino b t n ha o hdI t g forward muse b t postponed. I propose to substitute for it a brief statement of the results obtained in a kindred enterprise, the means for which were supplied throug e samhth e munificent benefaction.

Professor Haverfield lono suggesteag g d that informatio f reao n l importanc histore th r f Romaefo yo n coull probabilital dn i e yb secureorganisatiothe by d whaof n termehe t excavatorsan d ' 'flying column.' The idea was that a number of sites, on which the presence of e Romanth s suspectedwa s explorator,w mighfe a e visiteb td an yd cuttings mad eachn eo man n I . y case comparativelsa y brief examination would not improbably suffice to determine once for all the question of origin, while it was just conceivable that here and there a stroke of good luck might produce fairly definite evidence of date. Although the forminf o y difficultiewa g e sucth hn flyina i s g column r prove havfa o es d insuperable principlee th , underlyin suggestioe gth t int pu o s practicnwa e in July 1913 at the camp of Grlenmailen near , and again in followine th g summe came th Raedykef t po ra s near Stonehavene th n O . former occasion Professor Haverfield was fortunately able to be on the spot himseltako t d eleadina an f goperationse parth n i t . Fowore rth k at Raedykes I alone was responsible. In the discussion and description that follow t seemi , s preferabl deao et l with Raedykes first.

1 Proceedings, 1914-15 (vol. xlix.), pp. 93-138. 318 PROCEEDINGS OF THE SOCIETY, MAY 8, 1916.

I. RAEDYKES. This camp, whose name is variously spelt as Raedykes, Raedikes, Readykes, Re-dykes, and even Rhe-dykes, lies in the parish of Fetteresso

in , its defences enclosin1 g the greater part of the Garniehill or Garrison Hill, some three miles to the north of . movee Aon s s north through Strathmore e Grampiane rangth th , f o e s draws neare neared coaste th an r o .t r Beyond Stonehaved an a se e nth e mountainth s almost meet a considerabl thao r s , fo t e distance th e railway has to cling to the top of the cliffs in order to find a reasonably easy passage. Garrison Hill stands among the rolling uplands, about three miles back fro beache mviee th wTh . fro highese mth t point (628 feets i ) singularly picturesque, particularly toward weste sth , wher looke eon p su a, broad valley to the mountains of Deeside, or again towards the east, where the eye sweeps over the moor and up the long slope of the Kempstone Hil catco lt distane follod hth an wt a soutti sparklse e h th f eo to Stonehaven Bay. The surface of the hill itself is broken and irregular. In that respect the position is one which a Roman general would hardly have occupied except under stres f circumstancesso Romay An . n camp pitched there would emphatically have been one of those quce in loco necessario [constituuntur], uncle et necessaria castra dicuntur? We do not know 'who rediscovered it, but its earliest mention seems to be that in Maitland's History of Scotland, published in 1757. The description there given3 may be quoted in full:— " This camp is about three quarters of a mile square, or three miles in circumference, fenced with a high rampart and a very deep and broad gatese ditceac d th number n i f an h h,o ; x whic si fortifiee e ar , har d with a rampart and ditch, at the distance of about twenty-four yards without the said gates. This is the largest Roman camp I have seen, or can learn that there is in Scotland." Like so many eighteenth-century antiquaries, Maitland was hot on the scen f Mono t s Graupius confidens wa d earto t an ,bee d t n hthaha nru t i t believet laste a H . d that Raedyke Agricola's swa s camp thad an ,t " there is not the least room to doubt of this place's being the spot whereon the battle was fought." Just twenty years after the appearance of Maitland's History fine e thew th d n proprietor r RoberM , t Barcla f Urieyo , keenly interested s possibli t I . e tha s attentiohi t mattee s drawth nwa o nt r by certain chance discoveries made about this time in the immediate neighbourhood. A contemporary writer tells us that, when stones were Professor Watson, who mI hav e consulted, writes thae etymologth t s difficulti y e H . suggests1 , however, tha name possibly tth Englishee ema th e yb d forGaelia f mo c (an) Bath Dig?,, fosse-rathe th "earther o , n fort wit ditch.ha " Other variant Ree-dikee sar Bi-dikesd san . 2. .. Hyginus3 Mun.Vole D , . . 202i p .Castr., . . 56 THE ROMAN CAMP AT KAEDYKES. 319 being carted off for enclosing a field, " several urns were turned up," while "in a moss, hard by, two Koman hastce were found entire, and several others in a decayed state."1 However that may be, in the summer of 1777 Mr Barclay, who was a prominent public man and long member of Parliament for the county, carefully examined the site in company with General Robert Melvil d "hian ls respectable friend Lord Monboddo."

followine th n I g yea e ban th rf inquirer do s receive dfresa h recruin i t 2 the person of Mr John Stuart of Inchbreck, afterwards (1782) Professor of Greek at Marischall College, Aberdeen, who surveyed the ground and produced a plan of the enclosure.8 Finally, in 1784, the eccentric founder of our Society, Lord Buchan, took the subject up, doubtless at the instance of Mr Barclay, who seems also to have been responsible for the enlist- men f Professoo t r Stuart. As a result of all this activity we have quite a considerable body of ' literature' with Raedykes as its theme. It is noteworthy that none of the contributors accepted Maitland's view as to the enclosure being the veritable camp which Tacitus describes, an attitude of mind which is probabl e influence accounteth b y o b t yf Melvill o r e fo d , whose authority would naturally carry great weight. When this distinguished soldie d studenan r f militaro t y histor s broughywa t upo e scennth n ei 1777 s verdichi ,s quitwa t e uncompromising.4 "From every circum- stance," he came to the conclusion that the camp at Raedykes " could not be that occupied by Agricola's forces immediately before the battle." So sweeping a declaration might be interpreted as a refusal on the part of Melvil o entertait l e entrenchmen th e ide nf th o a t bein f Romago n workmanshi t leasa s i t t equalli t t allpa Bu y. possible tha tooe e th kth Roman origi f Raedykeo n r grantedfo s d tha s wordan ,hi t s meao n more e condescendthah s na ther l fa eventsy al o ssayt s upoA n .i , n details, his main objection seems to be to the theory that the actual battle took place on the : he " finds himself obliged to con- jecture, both fro remainse localite mth th d yan , tha conflice th t t there has been betwee e Danish th e Scotc th n d r otheo an ,h r northern invaders." Scottish lairds have always displaye t unnaturano da l reluctanco et relinquis e ownershith h e battlefielth f o p f Mono d s Graupiusd an , Mr Barclaexceptioo n rulee s continuee th H y wa . o nt believo dt e that e strugglth taked eha nKempstone placth n yield eo di de h e t HillBu . 1 Francis Douglas, A General Description of the East Coast of Scotland from Edinburgh to Cullen, etc., p. 261. The book was published in 1782, and the discovery is said to have taken place " some years ago." e Gough'Se 2 s editio f Camdeii'no s Britannia, vol. iii. (1790) . 416*p , , footnote accounn a r fo , f o t this visit.

See infra, p. 325. See Gough's Camden, I.e.

3 4 320 PROCEEDINGS OF THE SOCIETY, MAY 8, 1916. an important point to Melvill. He abandoned the notion that it was from Raedykes thae troopth t f Agricolo s d issueha ao measurt d e swords with the host of Gralgacus. Arduthy, two or three miles nearer Stonehaven, seemed to him to satisfy the conditions much better, and in e eighteentth h century there were still visible ther e remainth e f o s entrenchments, long since vanished f whico t t requirei hou , greao dn t effort of imagination to reconstruct a Roman camp. Raedykes retained s Romait n character s existencit t demandebu w , eno dfresa h justifica- thad di tr tionprovNo . e har discovero dt .Roma e "Th n general,s wa t "i suggested, " might, for various reasons, have been unabl pursuo et e eth advantages he had gained, and chosen to encamp upon the Garniohill, or Raedykes . . . an eminence which commands a prospect of the whole neighbourhood." The views just summarised were set forth at length in a communica- tion addresse Barclar M y db Loro y t d Buchan subsequentld an , y printed in Archceologia Scotica.1 Although the communication is undated, we know that it was read at a meeting of the Society on llth January 1785,2 and there is good reason to believe that it was penned but a few days previously. Some four "weeks later Lord Buchan basi e utiliseth s a t di secone lettero oth ftw f dso whic addressee hh Nicholso dt publishere th , , over the fictitious signature ' Albanicus.' Next year Nichols printed both in No. xxxvi. of the Bibliotheca Topographica Britannica3 under the heading, "Remarks on the Progress of the Roman Arms in Scotland, durin e Sixtgth h Campaig f Agricola.no e seconth n dI " letter, whics hi dated 8th February 1785, 'Albanicus' intimates that the local informa- tion which form s groundworit s reached kha afte m firss dhi hi r t letter written—thas wa , afteis t r 10th December mose 1784Th t. probable date for Barclay's communicatio s thereforni e December 1784 r thae fo , th t local information referre identicas i o t d l wit e communicatiohth n from Barcla doubtee yb wilt anyonno y ldocumento db placeo tw e wh sth s sid sidey eb . A larg e secon th par f o td lette f 'Albanicuso r n facti , ,is ' neither more nor less than a rechauff4 of what Barclay had said. The thesis maintaine s identicaldi . Indeed e verth , y languag s ofte ei e sameth n . The illustrations, however, are new. Barclay's communication to Lord Bucha beed nha n accompanie "ruda y db e sketch surroundine th f "o g country, which is here reproduced as fig. 1. An " exact plan " of the camp itself was promised, it being explained that in the meantime no proper measurements could be taken "on account of the depth of the snow." Th eAlbanicus' ' Bibliothecalettere th n si illustratee ar threy db e plates. The first of these is a map showing Scotland according to 'Richard of 1 . 565-70Volpp . i . . cit.,. Op 2 vol. iii., Appendix . 157p , . ff1 . . Pp 3

PROCEEDING E SOCIETY , 19168 TH Y F .O S MA , Cirencester,' which need not further concern us. The second is an improve slightld dan y extended editio f Barclay'no s "rude sketch,e "th improvement consisting in the substitution of what will turn out to be a fairly accurate outline of Baedykes for a very rough square. As this plate give gooa s d general ideae relativ t onlth no ,f yo e positiof no t furnishecampi e s s shapeth t a alsit f bu d ,o o valuabl,an a s ee cluth o et solutio problea f no shale mw l hav faco et e presently reproduces i t i , d here as fig. 2. The dotted line (A A A) represents the supposed route of the Roman soldiery from Strathmore to their camp at Arduthy (B), thence to the position (C) which they took up before the battle, and finall o Raedyket y s (D), where they reste n theido r laurels aftee th r victor won.s thire ywa Th d1 Bibliotheca plate wil referree b l latero dt . Meanwhile Professor Stuart's e discussiosharth n ei n briea call r fsfo beenoticehas n alreadAs . y stated mad,he firsehis t acquaintance with site 1778n th ei , when surveyee h e groundth dred e dan w plaa th f no camp opinioThe . arrivenhe then—adat n opinion whic confirmehwas d by many subsequent visits—was that Raedyke t Romano s t alla nswa . It was a "camp of the Caledonians . . . totally unlike those of the Roman Scotlandn i s , whic e universallhar y rectangular, wherea thin i s s one there is not a single right angle in its whole extent." Its attribution face awkwarth n "s t a o that wa e Romane t i toth du d s imitatioswa n of their mode of encampment, fortified with a wall and ditch, and having several gates with traverse n froni s f them.o t " This reads like th e languag n antiquara f o e whoyn i m there lingere traceo dn f Romao n fever. But, with it all, Stuart could not succeed in freeing himself from spele th f Mono l s unwillin a Graupius s wa s Barclae ga H . beed yha n to acquiesce in Melvill's criticism. For him, too, the Kempstone Hill remained the battlefield, and the fragmentary entrenchments at Arduthy were transfigured int e camoth f Agricolapo . Raedykes, howevers wa , surrendered to Galgacus. It became the camp in which the Caledonian chieftai deliveredd ha n t leasa r o t, composed memorabls hi , e oration. Such was Stuart's theory as expounded in his " Observations upon the Various Accounts of the Progress of the Roman Arms in Scotland, and of the Scene of the Great Battle between Agricola and Galgacus," published in Archceologia Scotica 1822.n i 2 This paper doe t seesno mhavo t e been read to the Society in the form in which it was printed. But it doubtless embodie e substancth s a communicatio f o e n entitled " Observationn o s some Remains of Roman Antiquity in the North of Scotland," which It would be mere waste of space to supply a complete 'key' to the lettering. PP, for instance1 , is 'the sea,' and UUU 'the Grampian Hills.' 2 Vol. ii. pp. 289 ff., reprinted in the author's posthumous Essays (Aberdeen, 1846), pp. 69 ff. Stuart's theor s recentlyha y been revive r CrabM y bdb Watt Mearns,(1914)e d K.C.Th Ol n f ,i ,o pp. 64 ff.

324 PROCEEDINGS OF THE SOCIETY, MAY 8, 1916. occupied two successive meetings in the winter of 1819 (Jan. 26 and Feb. 8), forty-one years after Stuart's original examination of Raedykes.

Barclay1 and Melvill were both dead, and Lord Buchaii had long ceased to interest himself activel e Society'th n yi s proceedings. papee s Th wa 2r thus a revival of an almost forgotten discussion. In the interval, how- ever e claith , f mRaedykeo a genuin o t s e classical ancestr d beeyha n fortified by its inclusion in General Roy's Military Antiquities of the Romans in North Britain, where a plan of it appears on plate L. s practicalli t I y certain himselthay neved Ro t ha f r seen Raedykes. He seem t eve no shavo nt e know existencs it f no e unti boos hi l k wass a , he thought, completely finished. There is no allusion to it in the text, and a scrutiny of the manuscript copy in the British Museum—which is more perfect tha n e librare Societ thath th n f i tf y o Antiquarieyo s of London—shows quite conclusivel e originaplaco th n n yi d e thalha t i t list of plates. I hope to set forth in detail elsewhere the evidence I have collected as to the circumstances attending the composition and the ultimate production of the Military Antiquities. Meanwhile it must suffice to say that the book was ready for the press in 1773, between which date and 1777 a few unimportant changes were introduced by the author. Thereafter the MS. lay untouched until it was posthumously published in 1793 at the expense of the London Society, one or two additional drawings having accumulate meantimee th n di . Fro language mth e used in their report to the Council of the Society by the Committee that had been appointe superviso dt issuee mighth e on , t naturally enough infer that this Committee had restricted their energies to the task of securing diplomatic accuracy of reproduction, and had scrupulously refrained from any attempt to edit. As a matter of fact, however, a close examination

reveals unmistakable3 indications of editorial handiwork. The editorial change f mosthao s i tt immediate interes s heru eo t t is the insertion of Raedykes in the Map of Roman Scotland,4 where it is accompanie descriptioe th y db n " Cast. Agricolce." woult I d seem that after finishine th t pu gd toucheh ha etext—thae th o st afte, is t r 1777—Rod yha e firsfoth r t tim eithed e e campha hear th d r f simultaneousldan o , r yo later on been able to secure a plan. Of this plan two copies were made, e bookMSSo th tw f .o e eacr th The fo f water-colouhe o e yar on r draw- ings, corresponding exactly in size and style with the illustrations originally included. Possibly the e froyar m Roy'brushn f ow snotI . , Arch. Scot., iii., Appendix, p. 172. 21 He resigned the Vice-Presidency in 1790 (Arch. Scot., iii., Appendix, p. 1). prefatore th n I y note that follow secone sth d title-pag Committee eth e say, quoting thein row report3 , "that it had been judged proper to publish the work from the manuscript, without any commentary, or deviation from the style and orthography of the original." Plate I. in the Military Antiquities. 4 THE ROMAN CAMP AT RAEDYKES. 325 they must have been executed unde personas hi r l supervision eithen I . r case their presenc e collectioth n i e s cleai n r intentionn prooa f o f o t revise the work once again, incorporating Raedykes. The action of the editors in reproducing the plan, and in inserting the camp on the Map, was thus fully justified. Nor did they go beyond their duty in adding the description " Cast. Agricolce," for Roy's own opinion was plainly expressed headline inth e which appear drawingse eacn th s o f ho : " Pla Agricola'f no s Camp calle Dykee dRa s neaabouy rUr Mile3 t s from Stone-Haven." One cannot help wondering how Raedykes was brought to Roy's very noticeyma welt I tha e .firse b l tth t information fro m cammhi o et Melvillt wil (i rememberede lb o wh , seed o came )ha ntw th e 1777n pi Th . wer acquaintancesd ol e thed an ,y must hav t froeme m tim timo et n ei London, where both were resident for a number of years before Roy's deat 1790n sucn hi O h. occasion e conversatioth s agaid n an mus nw no t have turned upon Roman antiquities, which formed the original bond of union between them, and it is difficult to believe that Raedykes was left unmentioned. Even if it were, however, so keen a topographer as Roy cannot possibly have missed the ' Albanicus' letters when they appeared in the Bibliotheca in 1786, especially as, in their more general aspect, they dealt with the very subject upon which the whole antiquarian world was just then expecting him to throw a flood of light.1 We may, therefore, conclud ee importanc thas alivth y 178 b wa o t tt elates a 6 y f Ro to e Raedykes min s madd hi Agricolanha ds p e u d tha wa ,an t ti . less i st I eas gueso yt s whenc whed ean procuree nh plans t dhi bu ; n analysi a e materiath f briny o s ma lgsolutio a mysterye th f e no Th . most convenient starting-poin r inquirou r fo yt quotatioa wil e b l n from Stuart, whose surve f 177s yalreado 8ha y been referre . Writinto d n gi Archceologia Scotica 1822n i Professoe th , r says:— "A drawing of this Scotish camp, originally made by the author in 1778, was sent by him to the late General Melville, an eminent antiquary, and early associate of General Roy ; and another furnished some years Eare afteth f Bucha o ly b rpublishe s uwa Nicholy db 36te th hn si numbe r of his Topographia Britannica. There is also an engraving of it, though meano n y b s accurate Generan i , l Boy's Military Antiquities." 2 From this passag cleas i t ei r that Stuart believe existence th n di f o e three distinct plan f Raedyke sdrawino n ow s hi f 1778g o s ) anothe) —(1 (2 ; r sent to Nichols by Lord Buchan about 1785; and (3) the original of Roy's plat whic, eL. h canno latee b t r than 1790 e yea th ,f Roy' o r s deathe Th . adverse criticism here passed upo e accurac th ne lasf thesth o t f yo e makes it certain that Roy's plan cannot have been copied from Stuart's Contemporary references sho wwels thawa l t knowti n that Roy's boo finishes k wa migh d dan t be 1published at any time. 2 Arch. Scot., . 301p . .ii 326 PROCEEDINGS OF THE SOCIETY, MAY 8, 1916.

drawing e otheth n r O hand. e absencsucy th , an h f criticiseo e th f mo Bibliotheca plan justifie assumptioe th s n l essential thaal n i t s thers ewa complete correspondence furthero g ' betweey Stuart'sd ma .an e t ni W . Fig show.3 thre e thire th sth ef d o plate s use illustrato dt lettere eth f so ' Albanicus.' It will be observed that it is entitled " Camp of Rhe Dykes 1785," as if the year mentioned were the one in which it had been executed. But in a postscript to the second letter it is described more explicitly as a " Plan of the Camp at Rae Dykes, on the Estate of Ury, in the Shire of Kincardine furnisheds ,a Robery b t Barclay of Ury, Esq.; from actualn a Survey, to Lieutenant General Melvill, in 1778."1 And the more explicit descriptio s amplni y confirme whay b d t Melvill himsel fs con sayhi -n i s tribution to Gough's Camden. After speaking of the visit he had paid to the spot in 1777, he proceeds:— "Mr Barclay having been pleased very obligingly e followinth n i , g year, to transmit to General Melvill a very accurate drawing, from an actual survey, of the camp called Raedykes, it was inserted in the XXXVIth Number of the Bibliotheca Topographica Britannica." 2 The sequence of events thus becomes plain. The " actual survey" of 177Stuart's—ws 8wa e know from himsel than i s ft thayeawa t i tr that it was carried out,—and the drawing which he made was subsequently forwarde o Melvilt d l through their common friend Barclay. Stuart's e Bibliothecath plad nan plan are, therefore t merelno , agreementn yi : the e identicalyar line kOn onl. s awantinyi completo gt e chaineth : we have stil discoveo t l r what part Lord Buchan playetranse th n di - action. This we are fortunately able to do. It is revealed in the following letter, writteEare th Nicholso t ly nb 3:— " To-morrow I shall send you the continuation of Albanicus's Remarks on the progress of the Roman arms in Scotland during the sixth cam- paign of Agricola, accompanied by a sketch of Richard of Cirencester's Itinerary a Topographica e countrd th an ,f o yp adjacene Ma l th o t t remain encampmentf so north-eastere th t sa nGrampiae pasth f so n hills.* Lieutenant-General Melville I find is possessed of a drawing of the camp at Rea-Dykes or Garnacaliill,6 described in the account given of it by Albanicus in his second letter to the printer of the Gentleman's Magazine, and I have written the enclosed letter to that brave, humane, and learned General, the rival of the Marquis de Bouille,6 who will probably permit you to use it for the purpose of rendering Albanicus's communication more satisfactory."

1 Bibl. Topograpli. Brit., xxxvi. p. 15. 2 Gough's ed Camden'f o . s Britannia, iii. (1790) . 416*p , , footnote. 3 Printed in Nichols's Illustrations of Literature, vi. p. 508. It appears under the date "1784 t havno ey beema nt writtebu n till earl 1785n yi .

4 These are respectively plates i. and ii. of Bibl. Topograpli. Brit., xxxvi. 5 A mere misprint for " Garniehill" or " Garrisonhill." So, too, " Garniohill," supra, p. 320. commandea 6 s Thaa , is t troopf Wese ro th n tsi Indies. •%„ A. The JVorfft Gate j4 ff M 'id*. T&.Tkefittft; Gats, jz nn-iJe. South*e C.Th Gate £t> wiJz. T^-T^u SaufJt West; (siite ,75 wide. E - ThMrtftWcst Gate ^wuk. F. A Cattoge lately built withi/i tfie Camp. G.The fann. TJT to fitive been- an i post.

H

Scale of Fttt. '•zoo 40a ooo Sao woo J>**&

Fig. 3. Plan of Kaedykes, from the Sibliotheca Topographica Britannica, No. xxxvi. 8 32 PROCEEDING E SOCIETY , 19168 TH Y F O S. MA , The identity of Stuart's drawing with the Bibliotheca plate being satisfactorily established, we may turn next to the engraving in Roy's Military Antiquities (plat emose L.)Th .t importan s showi t t pari f no t her e originas figea th , .4 l scale being maintained firse Th t . impression whic t leavei h s thaa countri s f o t y almost Alpin s ruggednessit n ei ,a picture ver r removeyfa d indeed fro reality.e mth fairnessn I 1 , however, t e i oughexplaineb o t t d thae engraveth t r alon s responsibli e r efo the precipitous character of the hills; no such abruptness of relief is suggeste e originath y db l water-colour sketch. This feature, therefore, may be disregarded. Concentrating attention upon the camp, one struce e cannodifferenb th t y kb bu t t manne whicn i r representes i ht i d here and in the Bibliotheca plate (fig. 3). Thus, fig. 3 shows only a single e easth gat t n i eside , while fig show.4 s two r compareO . , againe th , positio e gate th fign i f s no .wit3 E marke d he corre an th tha df C o -t sponding gate n figi s . .4 Small wonder that Stuart, with whoe mth drawincorrectnesn ow s hi g f woulo s d naturall n articla e yf b faitheo , felt impelle o condemdt n Roy's engravin meano n s y "a gb s accurate," especiall e confirmatioe lighth th f o tn i y n whic e supposehh d thas hi t own version received from the Bibliotheca plate, whose true origin he faile recogniseo t d . But the particular plate we have been discussing is not the only plan of the camp that appears in the Bibliotheca. The plate that immediately precedes it—illustrated here as fig. 2—is obviously a reproduction of the "Topographical Map" whic e Ear hth f Bucha o l n promise seno dt o dt Nichols along wit e seconth h Albanicus' d ' letter.thad An t 2wil e b l foun contaio dt nrepresentatioa f Raedykeno s which differ radicallo s s y from the larger plan of fig. 3 as to exclude the possibility of its having been based upon Stuart's drawing e otheth n lookee r b O hand .t i df i , at alongside of fig. 4, the resemblance to Roy's plan will be at once apparent. Indeed, when allowanc e differenc s madei th r e fo scalen i e , e agreementh t between remarkablo theto ms i othe y admio et an rf o t explanation than that both have been derived from a common source. This conclusio r quesou f t o appreciabln d bringen e th s y nearerf I . we can discover the origin of fig. 2, we shall know where Roy's plan ultimately came from. t Figwili , rememberede 2 .b l s merel,i improven ya d editio f figno , .1 e improvementh t consisting e mainlobviouth n i y s effort mado t e delineate the outline of the camp more precisely. Further, we have seen that Barclay was responsible for fig. 1, and that, in forwarding it to

impressioe 1Th n produce whole th y edb plat naturalls ei y much stronger than that produced necessarile th y b y limited portio t showi f no fign i . .4 2 See supra, p. 326. Fig . Pla4 . Raedykesf no , from Roy's Military Antiquities.

I 500 1000 1500ft. 330 PROCEEDING E SOCIETY , 19168 TH Y F O S. MA ,

Lord Buchan, apparently in response to a direct request, he took occa- sion to apologise for its imperfections, promising to let him have some- thing better when the weather conditions improved. His ipsissima verba are:— "I had the honour to acknowledge the receipt of your Lordship's letter. Since that time I have been at the camp of Raedykes, but could not take an exact measure of it, on account of the depth of the snow. As soon as the ground is clear, I will send to your Lordship an exact plan of it. I suppose, it contains about an hundred acres; but this is con- jecture. At present, I transmit to your Lordship a rude sketch of the country near it, which, if it can be understood, will convey the ideas I have formed."l This letter, as we have already learned, was written either in the end of Decembe e beginninth r 178n i r 4f o Januar o g y e surve1785.Th 2 y whic t hcontemplatei s would therefor made b e e latte e earlth n yri year; for snow never lies long so near the sea. Its outcome would certainly be ready in ample time to be inserted in Lord Buchan's " Topographical Map" (fig. 2), possibly not before that was first sent to Nichols on 8th February, but at all events before it was engraved for the Bibliotheca, t appeano wherd rdi tilt ei l 1786 t alongsidSe . f thiface o sth t thae tth small outline of Raedykes in the " Map " must be based upon an actual survey, since (as will be shown in the sequel) it reflects the reality a good deal more faithfully than does Stuart's drawing (fig. 3). What e inferencth s i e? Surely thad consequentl fign i an t , .2 Roy'n yi s plan, we have the result of a survey carried out by Mr Barclay's directions in e latth e winte r earlo r y sprin f 1785o g . Stuarte authon th a s f a o ,r earlier plan, would naturally hear from Mr Barclay of this fresh survey and of its immediate purpose. In the circumstances it is not at all surprising that he should have been led astray by the arbitrary insertion of the date "1785" in plate iii. of the Bibliotheca (fig. 3), and should hav planw e mistake ne drawin.n e ow th r s nghi fo It may be urged that Stuart ought to have been kept right by the postscript to the second ' Albanicus' letter, which gives a correct account of the provenance of the plate in question. Against this it must be borne mindn i , firstly writins , wa tha e 1822n gh ti , thirt r fortyo y years after e eventth ; and, secondly nevey , ma tha re havh t e seeBibliothecae th n itsel t all a knowledgs f ,hi e bein than gi t case derived fro summare mth y in Gough's Camden, wher e postscripth e s lesi t s prominent. George Chalmers apparently fell int e samoth e mistake, fo t seemi r s clear that s footnoteshi f figo o t tha , .t e e 2 refere th h ,tno on d n i sfigo an t , s .3 i t i circumstantialit f whicyo h affords convincing proofresa f o hf survey 1 Arch. Scot., vol. i. p. 565. supra,e Se 2 . 320p , wher indicates i t ei d that Decembe more th s eri probable. THE ROMAN CAMP AT RAEDYKES. 331 having been made for the benefit of 'Albanicus,' and at the same time supplies us with a detail of considerable interest by revealing the name of the person who carried it out. The words of the footnote are:— "Se Accounn ea Plaa f thid no san t Roman camp, fro n actuama l surve Georgy yb e Brown, land-surveyor e Biblth .n i ,Topograph . Brit., No. .36; Gough's Camdeii, v. iii. p. 416, pi. xxviii. ; Roy's Milit. Antiq., plaa e f thino se s d remarkablAn pi . .1 e groun e Transactionth n di s of" the Antiq. Soc. of Scotland, v. i. p. 565."1 Our analysiquita materiae o t th e s f unmistakabl u so s thu d ha l sle e conclusion. The ultimate source of the engraving in the Military Antiquities was the plan which Brown prepared, on Mr Barclay's instructions, early in 1785. Incidentally, this fully explains its superior accuracy f whico , e shalw h l hav n opportunita e f judgino ye th n i g sequel: a Browprofessiona s wa n l surveyor, whilen a Stuar s wa t amateur. How Roy obtained access to it, will probably always remain uncertain t seemi t sBu .obviou s tha t i must t have beem n hi sen o t t Barclar eithe LorM y b y r db ryo Buchan, possibly throug e agenchth y of Melvill, who knoe mw havo wt e bee n correspondencni e with both. It has already been pointed out that Roy cannot have missed the 'Albanicus' letters in the Bibliotheca. It may be conjectured that, in studying the illustrations, his trained eye was struck by the incon- sistencies betwee outline e cam nt th appeari th s f pa o e platn si e ii.—our fig. 2—an e morth d e elaborate plan- that occupies plate iii.—our fig. 3— and that he was prompted to institute inquiries. Such a hypothesis makes everything plain. Before we quit the older authorities, it may be useful to draw attention extraordinare th o t y differences betwee varioue nth s estimates that were formed regarding the superficial area of Raedykes. When every allow- ance has been made for the chance that some of the writers may be thinking of the Scots acre of 6084 square yards, and others of the English acre of 4840, the discrepancies remain sufficiently startling to serve as a warning against the too ready acceptance of eighteenth-century state- vanishew no e siz f th o ement do t entrenchmentss a s . Maitland's guess of "three-quarters of a mile square, or three miles in circumference,"2 seems to suggest a total of about 370 acres. At the opposite extreme is the next description to be published—that of Francis Douglas, issued in 1782,—according to which the camp "is an oblong square of twenty-one acres, has four outlets, with redoubts before them, and many of Caledonia, vol. i. p. 177, footnote (e). The careful distinction drawn here between a plan of

the1 camp and a plan of the ground makes it certain that it was plate iii. of Bibliotheca (i.e. flg. 3) which Chalmers believe reso dt t upon Brown's survey. Plat . (i.e.eii figs virtualli ) .2 y identical witgroune plae th hth f no d whic las e citee hth h t n sentencesi supra,e Se .. 322 p . * History of Scotland, vol. i. p. 202. 332 PROCEEDINGS OF THE SOCIETY, MAY 8, 1916.

trenchee th stile sar l pretty deep." Twelve years later Douglas's language

borrowewas d verbatim (withou1 t acknowledgmen sortthe any in ) of t notice of the parish of Fetteresso in Sir John Sinclair's Statistical Account* althougparise th s hwritere wa minister hth o ,wh , might have satisfied himself of its inaccuracy by half an hour's personal inspection or by five minutes' talk with Mr Barclay. Barclay's own estimate of the exten f Raedykeo t s comes neare e trutth e r th h f thao y n an tha f o t others. He puts it at "about a hundred acres." The Earl of Buchan

increases it to " about 120," while Stuart cuts it dow3 n to " forty or fifty." 5

The real figur ninety-thres ei halfa d .ean 4 The mai e rediscoverth n o factt s a s Raedyke f yo s havin beew gno n e e excavationn recordfre th o o ar turt e t e o t nw pu , f 1914o s n I . connection with these it is a pleasure to bear witness to the ready courtesy with which permission to dig was granted me by Mr Alexander Milne , Urie tenantfactoo th weretw o n estateso e r sewh th y welb s a , s a l immediately concerned Jamer M , s Burnett (Newbiggin Broomhilld gan ) and Mr Francis Gibson (South Raedykes). It should be added that I was specially fortunate in securing the services of Mr William Middleton, Stonehaven, as leader of our little working party. Such success as attended our operations was in large measure due to the keen personal interest which he took in the whole enterprise. Description of the results -will be facilitated if attention be drawn at the outset to a somewhat unusual feature that emerged at an early stage in our examination of the defences. A glanc t -Professoea r Stuart's plan (fig wil) .3 l show tha e liny th teb which the boundaries of the enclosure are represented is not of a uniform thickness throughout. Speaking generally mighe s i thay on , t sa i tt broad toward norte eastth sd han , narrow toward soute westd th s han . No such difference is apparent in Brown's plan as reproduced in Roy's Military Antiquities (fig. 4). Yet the variation which Stuart's drawing displays corresponds more or less accurately to a real and important distinction. Digging revealed the fact that the engineers who constructed the camp had made use of two well-marked types of ditch and rampart, choice th f typ eo eacr efo h particular section being obviously determined by the nature of the ground that happened to lie immediately in front. Wherever the conditions were at all favourable to an attacking party, the ditch assumed the character which we are accustomed to associate with fortification f Romao s n origin s V-shaped.wa t Thai , is ,t wita h widt abou f hdepta o feed 5 abouf 1 th o an t . Usuall7 t s facewa dt yi with

1 General Description Easte oth f Coast of Scotland, etc., . 260p . 2 Vol. xii 596. .p . s Arch_ Scot< vol L p. 565.

Bibliotheca Top. Brit., xxxvi. p. 15. Arch. Scot., vol. ii. p. 300.

4 5 THE ROMAN CAMP AT RAEDYKES. 333 puddle n bothe cuttingo e sidd Th r on o e.cla n yso which yieldeo dn evidenc f thieo s precaution were numben i , indeedjustifo t w s fe a r o ys , the supposition that its apparent absence was purely accidental, a consequence of natural decay. It seemed to have been applied with special care and elaboration at all points where the unevenness of the surface suggested the risk of a sudden rush of water after a heavy fall of thatd rain doubto An ,n . , supplie s ultimat it e clu th so t e e objecte On . instanc s particularlewa y striking t approachei s A . N.Ee th s . cornef o r the camp from the west, the ditch follows a route that could not but make naturaa t i l drain fo fairla r y extensive sectioareaa n I acrost . ncu a t si few feet befor e actuath e l turs reachednwa , scar d counter-scarpan p were firmle founb o t dy plastered wit ha laye f puddleo r d clay about 2 inches thick. This clay was hard—some of it so hard as to be almost like baked brick,—while much of it was of a peculiar red colour, quite unlike anything thae immediatth t e neighbourhood would furnish. Mr Burnett, the tenant, was of opinion that it must have been brought nearld frobe m a ymila e away. In the section of which we have been speaking the initial angle of descent of scarp and counter-scarp was maintained with approximate uniformity until the two met at the bottom and formed a V. Elsewhere e ditchth , when cleared out, presente dphenomenoa n closely analogous e Antonin th r Hil notee n Ba o l on t da eo t Wall. Abou inche9 t s above

the lowes1 t level, the two sides suddenly became perpendicular, the result being to leave a flat bottom, generally about a foot and a half wide, sometime vere s on yfullfeet o exceptionan I ytw . l cas e depte th eth f ho perpendicular inches8 s mtica 1 s s a h ,wa s whil e flae widtth th et f ho botto actualls mwa feety4 . This, however whers wa ,e ditc eth h abutted e roadwa e th edgo th nf o e y issuin e gatesth o f thags o , frote mon the conditions must be regarded as somewhat abnormal. Comparing Eaedyke noty s witer ma Hill thae hBa e variou,w th t s ditchee th t a s latter (which was, of course, a permanent fort) had an average width at f ratheo p to r e morth eaverag n thaa feed 6 n1 an t e dept f froho m 6-|o t - 7|; and that the perpendiculars at the bottom measured 18 inches, the flat space between them ranging from 2 feet to 8 inches in breadth. The genera shows l a effec s fign wa i t. 5 . Thus much for the first type of ditch. The second type, which occurred wherever the nature of the ground was such as to render a sudden onset exceedingly unlikely, was totally different in character. It was only 8| fee tcentrdepts e wide it seldo th s d n hi ewa an , m more tha nfooa d an t

1 See The Roman Forts on the Bar Hill, p. 28 (Proceedings, vol. xl. p. 430), where an English analog citeds y i . als e sectionCf oth e ditche. th f so t Newsteadsa , man f whicyo h displae yth same feature (Curie, A jRoman Frontier Post, p. 30). 334 PROCEEDING E SOCIETY , 19168 TH Y F .O S MA ,

shapee halfa Th . , too s distinctive,wa t borI .sor o en resemblancf o t e mereld ha t ybu bee , V n a mora scoope o n t r lesi eo t s dou perfunctor y manner. In other words, it was not a "fossa fastigata." Still less could it be classifie "fossaa s da punicata," othee on re varietth y whic Libere hth e d Munitionibus Castrorum s disposei recognise.o dt 1 Similarly s dimenit , - sions were only in partial conformity with the standard set up by Hyginus, Avho prescribes2 a minimum breadth of 5 feet and a minimum depth of 3, even for a ditch which is dug " loco securiori," as this was. These facts must be noted, but it would be a mistake to regard them as consti- tutin insuperabln ga e objectio viee th w o n t tha t Raedyke f Romao s swa n origin. After all e manua,th f Hyginu o lmoro n d e authoritsha y than y othean r practical text-book t i coul ;t fette no e discretiod th r f o n individxial commanders. s lesi st I eas o speat y k with confidence regardin e rampartth g , which has in most places been seriously destroyed. It seemed clear, however, that it also, like the ditch, was of two well- defined types. Wherever the ditc beed hha n deep e ramth , - part had apparently been high, d beeha n d e formean th f o d Fig Section. .5 , showing shape o Hillr f ditcBa . t ha eart he diggers throwth y b p n.u o substratun Ther s wa e f mo stones or of clay. Instead, a black layer, immediately above the original surface, indicated decayed vegetation, and proved that the excavated material had simply been tossed on to the grass or heather as it grew. It is impossible to give any estimate of the original size of the mound. Wher s besi t eti preserved s greatesit , t height to-da s abouyi t 4J feet, whil e blaceth k layer extends continuously backward distanca r sfo f eo 19 or 20 feet from the inner lip of the ditch. The figure just mentioned cannot, however e assumeb , o represent d e actuath t l breadte th f o h bas moune efor; th s ,a d crumbled e blacth , k layer would inevitably tend to spread. It should be added that there appears to have been no berm, the outer face of the rampart rising in almost a direct line with the inner face of the ditch. Turnin thaty e seconsa th ,y o gwherevet d ma type e e ditcon , th r h became narrow available ,th e evidence suggested that ther beed eha n a corresponding chang charactee th barriee n ei th f ro r begibehino T n . dit with, wher bess i mort o ei t n preserveew thano s feeni 4 t r 3di tJo high, and this may be a somewhat more reliable index to the original size than 1 Op. cit., c. 49. a Ibid. THE ROMAN CAMP AT RAEDYKES. 335 was the 4£ feet of the earthen mound, for the material used had consisted of large, loose boulders, evidently gathered from the hillside, so that the natural process of disintegration would be relatively less rapid. Again, the apparent breadth at the base was only about 8|- feet, as compared probabls i t i d her r 20wito o an ;eto 9 y h1 unnecessar mako yt e much allowance for " spread." Even so, however, the second type of rampart, like the second type of ditch, must have fallen distinctly short of the requirements formulated by Hyginus, who lays down a minimum breadth of 8 feet and a minimum height of 6 as suitable for a vallum. That, 1 however, was " loco suspectiori," and it is, therefore, hardly a fair test. We pass nexmora o t e particular descriptio came wholea th s f pa no . The irregularit outlins it pronouncef o yo s s ei d that ther geometricao n es i l term which could convey an accurate idea of its general configuration. Tha onln e gaineca tyb referriny db g directl e plath no yt (fig r . 6).Fo 2 our immediate purpos t wil ei conveniene b l treao t anglee t th fou f o sr (P, Q, E, and S) as principal angles, and therefore to regard the enclosure itself as a quadrilateral. Nor is it only considerations of convenience that suggest suc hcoursa desirables ea f attentioI . positioe paie th n b o dt f no the six gateways, as indicated on the plan by the first six letters of the alphabet t wili , observee b l d that they fall naturally into three groups, each of which may not unfairly be called a pair of opposites—A and E, F. Onc eand thiC s D, fac beehas Btn and realised needit , greasno t effort of imagination to see in Raedykes a Roman camp of ordinary form, wit s sidehit s deflected from their norma n le resul a th line f s o a ts endeavour to accommodate themselves to the sinuosities of the ground. Thi poina s s i whic o t h we.shall have occasio returno nt . The north-eastern came anglth pf e o projects lik hugea e salient into e moor botth n O h.salien e sideth ditce f o s suffereth t hha d compara- tively little fro neglece mth f centurieso t , while considerable stretchef so the rampart remain in fairly good preservation. Although the actual s beecorneha n) partiallr(P y obliterate farm-roaa y db d running north soutd an h acros e defencessth , s lefenougi shoo t t i w d f ho thaha t i t originally been rounded in the usual Roman manner. The general con- ditions, in short, seemed to indicate this as a suitable spot for beginning our excavations t therefori d an , e become naturala s starting-point foa r statement of their results. The ground along which the eastern side of the defences (P Q) runs slopes gently from north to south. Immediately in front is a stretch of flat, open moorland, toward e souther th f swhic o d hnen there risea s

1 Lib. de Mun. Castr., c. 50. In the preparation of the plan I have received very valuable assistance from Mr J. Mathieson of th2 e Ordnance Survey Department. 6 33 PROCEEDING E SOCIETYTH 8Y E , O S1916 MA , .

hillsw groulo , f havinpo g betwee e camn th the pd n i mwhaan s wa t earlier times apparently a bog. The total length of the line from P to Q is about 2626 feet, inclusive of the gaps for entrances at A and B.1 r abouFo t three-quarter f thao s t distanc s virtualli t i e y straight. Immediately beyond gate B, however t i ,swing s slightle eastth ,o t y continued an folloo st alteree wth d direction untireacheds i Q l . From P to gate B it forms the boundary between the farm and the unculti- vated moor. Here, accordingly, the ditch remains distinctly visible, while (as already indicated) there are still extensive traces of the rampart, especially betwee gatd . an eWitA nP change hth f directioeo n comes another change: not only is the rampart completely levelled, but even the course of the ditch can no longer be readily determined. The ex- planation lies partlface th t n ytha i e grounth t d herew , thougno s ha t hi reverted to moorland, has at one time been under cultivation.2 But ther alss ei ofurthea ditce gatfroo y th t hmeB P wa r e reasonth l Al . has been of the deeper or more formidable type. It has continued to be so for som yard0 e6 s beyond gat towardeB . ThereaftesQ rapidlys ha t i r , grown shallower, soon merging into the second or slighter type described above. Therevidenco n s ei shoo et w whethe e ramparth r t underwent lika e transformation naturas i t i t supposo t l bu , e tha t didi t . The distanc s showe 25-inca th , e Q n frono ho t mOrdnancP f o p ema 1903, approximates very closely to the measurement given above. Exact compariso , howevernis betweeo tw e , impossiblenth , partly becausf o e difficulte th f iixinyo identican ga l starting-poin partld an , yP becaust ta e e Ordnanc th s i avowedl n p o e positio th ma Q e y f o nconjectural . Besides, the officers of the Survey have taken no notice of the gates, and their reason for ignoring them is plain. Although gaps in ditch and rampart are apparent both at A and at B,3 there is nothing on the 1 It may be interesting to compare the details of this measurement with the surveys of Brown (1785 Stuard )an t (1778) s recordea , e platee Militaryth th f n di o s Antiquities e Biblio-th f o d an theca respectively. Owing to the smallness of the scale on which they are reproduced, the margin of possible erro connection i r n with these, particularl lattere yth dangee s largef i ,th o d , ris an ; course, specially great where the distances concerned are relatively minute, as is the case with s furthe gatesspacerememberee e e ha b th th t r I o t r.s fo d that Stuart faile notico t d e gatt a eB all. The following table brings out the facts, the distances at the extremities being reckoned from the middle of the ditch :— Macdpnald. Brown. Stuart. From P to A = 917' 840' 930' Gap at A = 64' 110' 52' = B Fro o t m938A ' 910' -| = GaB t 41pa ' 90' V1480' = Q Fro o t m666B ' J 660' 2626' 2610' 2462' 2 Probably durin firse gth t fift sixtr o y y yearnineteente th f so h century. 3 That at -B is much more indefinite, owing to the greater extent of damage done. The dis- tances from P and Q and from one another are given in footnote 1, supra. THE ROMAN CAMP AT RAEDYKES. 337 surfac mako t e e their purpose clear. Since 1785, whe plae nth n repro- duced in the Military Antiquities was made, the tutuli or covering ditches have entirely vanished. They were easily recovered wit assistance hth e spadee o th fwerd an , e examined with some care particulaA . r descrip- tio f theno m seems desirable. e ramparth Th e cannoA b n e i t w exaca tp no tga t e widtth f o h determined. But the width of the gap in the ditch was 64 feet. The interva coveres wa l d with puddled clay, which sloped gently down into ditce th eithen ho r hand distanca feet 8 a 3 . tf d Opposito ean , it o et fro outee e innem th e ditchth rth s line r e f edgth edgeo ,wa th f f eo eo tutulus. Thi fees 1 latte6 t s longr wa fee 1 ,1 t broad fee5 d t deepan , . Both coverina d ha side f puddled go ha s d clay. They sloped inwards until within abou bottom9 inchetthe of s , when they suddenly became perpendicular, formin trenca g h abou a hal a food ft an e widet Th . analogy with the ditch of the main camp does not need to be emphasised, but it should be recorded that, about a foot and a half from the bottom, lina f decayeeo d vegetable matte f considerablo r e thicknes noteds swa , indicating perhaps the depth to which the tutulus had been open in 1785, subsequen whico t t hs probabl datwa t t epurposi se y f filleo ep u dwhe n the ground was put under cultivation. The tutulus at gate B (which was almost directly opposit e farmhousth e f Broomhillo e ) provee b o t d similar in shape and construction, while its distance in front was exactly the same, 38 feet. It was 12£ feet broad and 4 feet 8 inches deep, with moro clan s ey thaupowa feesidest e i 9 n n3 th t t longbu ; figur,a e which perhap feetsditce 1 4 justifieth .t ha estimatinn n i i Fro s p u sm ga e gth the fact that no layer of black mould was observable, we may conclude that this tutulus filles verp wa du y earlyobvious i t I . barels thawa t yti visible in the eighteenth century; for, although Maitland and Brown detected it, Douglas and Stuart missed it completely. As they also

missed the gate itself, the inference is that, at this point1 , rampart and ditch were the mucn ni e samhth e conditio thas na whicn i t fine hw d them to-day, the ruin being so considerable that the gap for entrance longeo n s i r properly distinguishable cornee s cleareTh .t Qa rwa t dou with the spade, and proved to be rounded as P had been. The stretc representR h o frot mQ e shortes e th fous th rf o sidest . It is also the side which is most difficult to trace. As a rule, it is accessible, from the outside, only after climbing a long slope, with e resulth t tha e slighteth t r typ f defenceo s beeha en deemed adequate throughout ploughe Th . , therefore ,perio e whicon r anothet do ha s ha r been busy over all save a fraction of its length, has found the task r r StuarMaitlan above 4 1fo Fo Browd d d an Douglad te figse an an ; th an d se n3 . e se s 331d an . quotation8 31 . pp n so VOL. L. 22 338 PROCEEDINGS OF THE SOCIETY, MAY 8, 1916. of obliteration fairly easy. Except in two short sections, nothing what- eve s visibli r e upo e surfacenth . This lac f obviouko s evidence finds it s reflection in the differences between the surveys. According to Brown the length of Q R is only 960 feet, while according to Stuart it is as much as 1380. As the two are in general agreement regarding the plais i positio t i n , tha Q disturbin e f ntth o g factor uncertaie musth e tb n e measurementTh positio . R f o n s take n 1914i n , base s theda y were upon investigatio e spadeproveo th t y o nb g , that Stuar approxis wa t - mately right: they giv totaea l distanc f someo e 1300 feet betweed an nQ R. At the same time they furnish a convincing explanation of the error into which Brown fell t willI . , therefore e convenientb , mako t ' e th e record somewhat detailed. The poin fee lie0 Q t st11 well-markeeasa f o t d angle walth ln e i form - eastere th g in n boundar farm-roae th f yo d leadin Broomhillo gt s it n I . progress towarde defence th e lin t f th firsa e o s R i ts entirely obliter- , howeverby e d ditce th tracan th h ,f ated y o kbecomeB . s quite apparent, and behind it runs a series of boulders representing the remains ramparte oth f . These trace firse sar t noticeable abou fee8 t8 t froe mth spot wher e wal th es intersectedi l , this spot s turbeinit nn i gabou t 170 feet south of the angle mentioned above. So soon as the wall is crossed the clue disappears entirely. The line is lost beneath the farm-road, and it fails to emerge in the cultivated land beyond. Digging, however, showed tha t continuei t almosn ru o dt t straigh founs twa dont I that. , after traversing diagonally the corner of the first field (O.S., -1903, No. 2301), the ditch presently enters a second (O.S., 1903, No. 2302), the boundary between the two being crossed 65 feet out from the western margin of the farm-road. A short stretch of 24 feet then led to the gap at gate C. This gap turned out to be 58 feet wide. About 20 feet in tutulus,a y la fee front 6 i 5 f t o tlong t otherwis,bu e presenting exactle yth same characteristic wers a s e e maishowth ny b nditc h throughoue th t whole length of the side Q R. That is, it was about 8J feet wide, and about a foot and a half deep in the centre. It is clear that it can never have been intende serioua e b o dst obstacle thad attacn ,an ta k from this quarte regardelass e th trwa degren i s da e improbable. Beyond gate C the course already set was pursued with little or no deviation. fee0 Fot26 rr ther o som 0 s nothinei e25 e seenb o gt . Then e ramparth t suddenly reappears, shortly afte e s linreacheth ha re a d rough cornee fielth df o wherr e tiltmoorland han d merg inte e eoon th other without any wall to divide them. It remains in good condition for about 190 feet, only stopping short a foot or two on the hither side of a third field (O.S., 1903, No. 2304), beyond the boundary wall of which the surface once more resumes its normal aspect. The.boulders used to form E ROMATH N CAM9 33 T RAEDYKESA P . the barrier are here exceptionally large, some of them being from 2| to fee3 t long, with their other dimension proportionn i s doubo N e . th t difficult f movino y g such unwieldy blocks account r theifo s r survival n situ,i their fate formin n thii g s respec a curiout s contraste th o t obliteratio f theino r companion ditch presence th , f whiceo fronn hi d ha t to be verified by digging. We must suppose that elsewhere along Q R the rampar s beeha t n deliberately destroyed y organiseb , d e effortth n i , interests of cultivation. At the farther extremity of the section we have been displace,e describingth f o w d fe stone ,a s have been thrown inte oth moor behind, givin spoe gth t somethin appearance th f go cornera f eo . This last featur withouis e t doubt responsibl mosthe t for eseriou s blemish that disfigure e generath s l accurac e plath nf yo reproduce n i d Eoy's Military Antiquities (fig. 4). A comparison of measurements makes it clear that Brown took it for granted that the line swung to the right herd proceedean e d straight s througwa e e mooH th ho gat t . r eD obviously misled by the displaced stones that have just been referred to, perhaps because in carrying out his survey he approached the spot from the north-east, precisely as we have done. Had he come to gate D first, a careful search in its neighbourhood would probably have given him a hint that would have enabled him to avoid the mistake into which he actually fell. Digging amply confirmed Stuart's diagnosis. Instead of swingin poin e righe th questionth n t i t a tstraigho gn t o line n th ,era t ahead, passing throug e cornee thirhth th f do r field (O.S., 1903 2304). No , , crossin gfourta h (O.S., 1903 . 2308) d finall,No an , y enterin gfifth—tha e steeply sloping field immediately above the farmhouse of Garrisonhill (O.S., 1903, No. 2307)—about 109 feet from its eastern edge. Within the norther e last-namen th hal f o f d fiele pointh d undoubtedlR t ye liesTh . e cropstatth f so e preclude endeavouy s exacit dan x tfi positioo t r n by excavation. On the other hand, it was easy to determine it approximately, by conjecture, as the point of intersection of two adjacent sides whose general directio knowns distancne wa Th . e fro nearese mth t side of the gap at gate C must have been about 832 feet, giving a total length of about 1300 feet for the side Q R.1 AvashortesR e th sQ f mosd I an t t regula foue th rf ro sides whil, S ,R e not the longest, was certainly the most irregular. The ground which it The following comparison is on the lines of that made in footnote 1 on p. 336:— 1 Macdonald. Brown. Stuart. = C Fro o t m410Q ' 370' 76^ Gap at C = 58' 90' 60' From C to R = 832' 500' 560' 1300' 960' 1380' Note that, while Stuart's tota approximatels i l y right (allowance bein ge difficultie madth r efo s mentioned in the former footnote), he blunders seriously regarding the position of the gate. 340 PROCEEDINGS OF THE SOCIETY, MAY 8, 1916. traverses is at once .undulating and high, with a steep fall towards the outside. Henc otherwise eth e unintelligible turns whic line hth e displays as laid out upon the plan, and hence also the use of the less formidable type of defence for some six-sevenths of the entire distance. From the angle R (which we .may safely assume to have been rounded) as far as gate D the course followed appears to have been straight. The shallow ditch was recovered by digging at three intervening points:—firstly, where it returns to field No. 2308, about 340 feet west of the point at whicentered ha t hi d it; secondly, wher t passeei s from 230fiel. 8dNo into fiel2309. dNo , abou fee0 t6 t fro e northermth n extremit wale th lf ythao t separates the two; and thirdly, where it once more reaches the open moor, about 118 feet west of the extremity just spoken of. Even on the moor the track is at first extremely hard to pick up. For 145 feet the surface indications are of the faintest. Still, a close scrutiny will detect probabls wa them,t i yd theian 1 r discovery that enabled Stuar steeo t t r clea f Brown'ro s mistake tha d mads an , ha t t epossibl i officere th r f efo o s e Ordnancth e Surve reconstruco yt e outlinth t f thio e s portioe th f no camp with suc hneaa r approac accuracyhto . At the end of this almost obliterated stretch of 145 feet comes the gap at gate D. The break in the ditch was 45 feet long.2 At a distance of some 30 feet in front of it was a tutulus, which resembled in general characte e correspondinth r e gth defencf o s t gata e wa . t ThaeC i , is t same breadth and depth as the shallow type of ditch, clearly because the nature of the ground was here of itself sufficient to provide immunity against seriou disappearew s attackno resulta s s ha A t .d i , entirelyt I . was, however, recovered by excavation, when it proved to be 33 feet long, or 12 feet less than the opening which it was supposed to cover. For

rather more than 1000 feet beyond the gap no difficulty whatever present3 s itself. After an initial inclination towards the right, ditch and rampart, fac e spitn i th t f ethao t they hav eslightee beeth f no r type, remain con- spicuou r nearlfo s yard 0 y 30 the s salmosa n y ru straigh a n ti t line across the shoulder of the hill. Boulders peeping out from the overgrowth of whin and heather serve to show how compact has been the structure of the vallum. When the steep slope, now occupied by the cultivated fields that lie around the deserted homestead of Mid-Raedykes, is well within sight, there is a sudden swing to the left, no very obvious explanation of which can be suggested. Just where the descent begins to become pro- nounced, ther indicatione ear attempn a f so removo t ramparte eth e th : 1 In 1914 they were almost wholly obscured by whins. On revisiting the camp in 1916,1 found thae whin th td bee ha sn e burnee intervaltas th th f followin o k n d i d an , s e linth gwa e consequently somewhat easier. 2 This is Stuart's figure. Possibly it may have been rather less, seeing that the covering tutulus onls fee3 wa y 3 t long. precedine se t Bu 3 g footnote. E ROMATH N CAM T RAEDYKESA P . 341 boulders are scattered about in a confused way, as if an unsuccessful endeavour had been made to clear the ground for agricultural purposes. Fortunately, the ditch maintains the clue intact. Recovering something like its former direction as it passes through the corner of the first enclosure (O.S. ,. 2284) 1903No t ,i enter , e seconth s d (O.S., 1903 . 2283No , ) abou fee0 t4 tdividin e beloth f uppee o w th d g linren e between them. The section that immediately follows turned out to be more difficult to trace than any other, and in the end the hope of discovering gate E had to be reluctantly abandoned. This is scarcely matter for surprise, since it is plain from fig. 3 that even in 1778 the defences at this point had almost completely vanished. Digging in 1914 showed that Stuart's con- ception of their course was fairly sound—a good deal sounder than Brown's, although his measurements were much less accurate. Fig. 4 represents them by a broad straight line running direct to gate E from the point that was reached at the close of the preceding paragraph. Fig. 3, on the other hand, ventures onl n faind somewhao y an t t sinuous markings. The spade revealed ample justification both for the faintness and for the sinuosity. All the way through field No. 2283 the ditch continues to be slightee oth f r typepeculiarls thao wa s , t ti l fitteyil resiso dt erosive tth e forces that would naturally t attactraversei s a stee e t k i fac th e f pdth eo slope and dropped gradually down to the level. Consequently the traces it has left are very indefinite. It was, however, possible to make out that crossed ha fiele t i dth d diagonally, wit t leas ha distince on t t deviation fro straighte finallyd mth ha d t immediateli passean ,f o t dou y above eth north-west corner. By this time it has reached the level, where further investigatio barres ninterpositioe i th y db farm-roade th f no . t seemei Eve d nha d pruden tampeo t r wit roadwaye hth chancee th , s of a successful search beneath it would have been almost infinitesimal; e slighteth r typ f ditco e h could hardly have survive e extensivth d e ' making up' to which the ground has been subjected. Here, therefore, recourse must be had to conjecture. A firm basis for it will be found by crossin e roapoina gth o dt t directly opposit e deserte e dooth eth f o r d farmhouse. Nothin s e visibl surfacgi e fiel th th n df eo o e beyond t Bu . trenching showed that it was just here that the ditch emerged, and showed, too, that when it did emerge it was of the deeper and more formidable type. Ther s thu e i t doubgato a n s liep eE tsga thaundee th e t th r farm-road tha d figs seem3 an ,.(a t t thithaindicatep o a t s s sga t wa t i ) e characteth e defenceth f o r s changed e changreasoe th Th r s .i en fo self-evident terraie th : n that lie fronn si weld flas i t an tl adapted for the massing of an attack. Leavin e moments follo y a th e gat r r e gditcth ma w fa efo th e s ha w , the corner S. No surface-signs were available for guidance; but by the 342 PROCEEDINGS OF THE SOCIETY, MAY 8, 1916.

aid of cuttings it was traced through the field in front of the farmhouse (O.S., 1903, No. 2279), about 80 feet from the north-east angle of which it passes into the next field (O.S., 1903, No. 2258). The point S lies about 158 feet beyond the dividing wall, or about 360 feet from the spot where e ditcth h reappear e road e edgth th f . eo t sa Curiousl y enoughe w f i , tur shalnearle s thafeefige no a 0 w t s se 4 .li t t 36 possibl s ya distance eth e by which S is there represented as being separated from the gap at gate e thuE.W 1s reac anothey hb r rout e conclusioeth d t whicna ha e hw already arrived as to the site of gate E being buried beneath the farm- road. In all probability the tutulus is similarly concealed. At all events, searc t elsewheri r hfo fruitlesss ewa therd an ,abundan s ei t e rooth mn i suggested hiding-place t musI .born e b t minn ei d thatutuluse , th t E t a havy whilma e t ebeei n deep exceptionalls wa , y short gatewae ,th y there being very narrow, much narrower than any of the others. This we know from feetwhosStuart30 and ,give at e who ,it srecord , beina g e morth l verbae al confidentl e b l y one,ma y relied upon becauss it f o e

comparativ2 e accurac regarn yi otheo dt r gates. The condition of the crops unfortunately prevented the actual location of the corner at S by digging. But, in the light of the two eighteenth- century surveys, its position can safely be determined by producing the two adjacent sides till they meet. Equally, the general trustworthiness of Brown's plan justifies us in assuming that the angle was rounded, just as P, Q, and presumably also R, had been. It was apparently rather larger than a right angle. For some distance after quitting it the line of the ditch has been completely obliterated. Presumably, however, it continued to be of the deeper or more formidable type. That is what the nature of e grounth d requires whad an , t Stuart's plan suggests. Moreovere th , presumption becomes a certainty when the surface-indications appear again, as they do about 360 feet beyond S, as the line begins to ascend an uncultivated slope. Some 160 feet farther on, the terrain in front once more becomes difficulassaultinn a r fo t g party, wit resule hth t thae tth characte ditce th hf ro promptl y changes, continuing shallo stretca r wfo h of some 550 feet—that is, until the gap at gate F, which stands on the hille bros beeth ,ha f wno reached alteratioe Th . evidens ni t enougo ht the eye. But to make assurance doubly sure, it was verified by excavation. Rampar ditcd tan h have bee ngooa d deal disturbe immediate th n di e

Stuart make t aboui s t 600 feet, wherea outside hardln th t ca sa t ei y have been more tha0 n38 feet1 or 400 feet. He is thus nearly as far astray here as he was at gate C. In view of the doubt as positioe th o t gatf t doen i seeo t , eE sno m worth while givin gcomparativa e tabl measurementf eo s side foth re mai e BSTh n. differences wil apparene lb t fro texte mth . plae sidnotee e th nf th th e o reproducet e sa Se figs da . 3 Thi. s explains comparae whyth n i , - tive2 tables, Stuart's reckoning for the gates is always so much more accurate than Brown's, whose real estimate cannot fairl judgee yb d from suc hsmall-scala e pla figs na . .4 E ROMATH N CAM3 34 T RAEDYKES.A P

neighbourhood of gate F. It was, therefore, impossible to ascertain the precise width of the gap. Stuart, whose information on the point has some appearance of precision, makes it 74 feet. On • the other hand, the now filled-up tutulus, which lay about 28 feet in front, was cleared out and shown to have been no more than 56 feet long. With a breadth of depta 1 0d f 5£feetho ha , t whili , e sideeth s slope withio dt inchen9 f so e bottoth m when they became perpendicula reacheo s d lowese an rdth t leve hala food la f an apartt . These dimensions show tha e defenceth t s are entering a fresh danger zone. When ditch and rampart begin on the farther sid f gateo , the eF e onc yar e e moragaith f eno formidable type, and this persist defencee se corne withouTh th s . a P rs r breaa t fa s ka stron e need b ha o dt g here r the,fo y actually faced upwards towardsa gentle incline. To-day they serve as a boundary between field and field, and consequently they are fairly well preserved for virtually the whole of the 1071 feet included in this section. The rampart is seldom less than 3 feet high on the inner side, and it rises gradually until it reaches culminatina spoe th tt a fro e gbac w mar poin k no e whict P.W a t1 h we started, and have next to consider what historical deductions, if any, we shall be justified in drawing. Wherever the ground was opened, the earth was turned over very carefully hope th en i ,tha t mighti t yield some coi r somno e fragmenf o t pottery which would histore e camp servth clua e th s o f Th et a ey o . result was disappointing. The only object found was a shapeless mass of iron, whic s recoverehwa d fro e botto me ditcth sectio a th n f mhi t o ncu across it a little distance to the eastward of gate F. The subsequent application of preservative treatment brought to light certain features which suggested that this mass may once have formed part of the hub of curious i t i wheel a d s thaAn . t wheels already figure prominentle th n yi very meagre list of 'finds' that have been recorded from the area of Eaedykes. Professor Stuart r instancefo , , writin n 1822i g , says that "fea w years since" ther beeditcethe had n h take of " smala nout l hoop or ring of iron, of the rudest workmanship, and much corroded, being about four inche n diameteri s d veran ,y thick, which coule b d imagined useful for no other purpose than to contain the axle of one of their war chariots."2 It is still to be seen in the Museum of King's completeare sidP the datS e the I giv, a for 1As ecomparativa e tabl measurementsof e , corresponding to those already given for P Q and Q R:— Macdonald. Brown. Stuart. From S to F = 1070' 1040' 1060' Gapat, = F 58'(?) 80' 74' = FroP o t m1071F ' 108(X 1200' 2199' 2200' 2334' 2 Arch. Scot,, ii. p. 301. 344 PROCEEDINGS OF THE SOCIETY, MAY 8, 1916.

College, Aberdeen. Again, some time between 1822 and 1845, when

the New Statistical1 Account was published, a complete wheel was dug up within the enclosure and transferred to .2 A special casconstructes ewa precautiono n contaio dt t bu , nit s were taken against decay and corrosion. To-day its remains are represented by three portions of the iron tyre, sadly rusted but still about an inch and a half broad. To judge by the dimensions of the case, the original diamete beed ha rn full feet3 y . It is by no means impossible that the three wheels, whose existence is thus indicated y havma e, belonge e Romath o t dn period; wheels n elaboratoa f highld ean y finished type have been discovere morn do e than one Roman site in Scotland. But even the acceptance of such

a view woul t carrno d s ver yu y far, sincs alwayi t i e o t s s opeu o nt 3 t morge attribut e w substantia o d e e r Celtsthemth No o lt . help from the vague statements that have come down to us as to the finding of other objects at Raedykes. The "urns" of which Francis Douglas spoke were almost certainly native. And the rest of the evidence

e bringh s 4 forwar s equalli d y ope o questiont n . Whe e speakh n f o s "Roman hastce" "d a an Roma n spear," s usini e adjective gh th 5 e after e loosd th unscientifian e c manner whic n s customarow wa h s hi n i y day, when antiquities of the bronze age were almost universally supposed to be Roman. Indeed, the case is unwittingly given away by Professor Stuart, who, referring doubtless to the very objects that Douglas had in view, mentions "heads of spears of mixed brass, as almost all those in Scotland ascribe e Romanth o dt smore are."Th e circumstantial report

of Lord Buchan amount littlo t s6 e more, althoug knowe hmattew a s a , r of fact, that the two last among the articles he enumerates were of iron. He says:— '' Several Roman weapons have been found in this camp, particularly a hasta helmetd an f whice lawyerso , th e forme hth n i s i ' rlibrar t ya Edinburgh ; and lately a fragment of another hasta and a malleolus have been dug up." ' As there are no associated objects which throw a clear light upon the origi Raedykesf no have w , e perforc falo et l back upo e testimonnth f yo 1 Proceedings, xxii. (1888), p. 358, where the diameter is given as 8 inches. Doubtless this refers to the outer diameter, and Stuart's measurement to the inner. 2 N.S.A., vol . Kincardineshire,.xi . 250p .

Proceedings, xl. ff.(1905-6)4 , 49 wher . pp ,e possibility eth Celtia f o c origi s discussedni . Also Curie3 , A Koman Frontier Post, pp. 292 ff. 4 See the passage quoted supra, p. 319. 5 General Description, etc., p. 261. B Arch. Scot., ii. p. 301. 7 Nichols's Bibliotheca Topographica Sritannica, malleoluse e Th th . xxxvi. d No 13 an . p . fragment of the hasta, which were of iron, were presented to the Society of Antiquaries of Scotland (Arch. Scot., iii., Appendi t the. 38)p Bu . y.x ii see havmo t e long since perished.

E ROMATH N CAM5 34 T RAEDYKESA P . the entrenchments themselves. And here the eye is at once caught by a feature tha highls i t y suggestiv f Romaeo n methods. Scotland offero n s exampl cama f f demonstrableo po y native construction s whicit s hha entrances protected by traverses. In this case, however, each of the six gates has in front of it a tutulus such as the Roman military manuals prescribe e foun ar d suc indubitabln s an i ,da h y Roman forts like eth Antonine fort at Bar Hill. It may be argued that there was nothing to

preven e Caledonianth t s from adoptin1 g devices whice us hn i thew ysa among the invaders: fas est et ab hoste doceri. The possibility must be allowed, but the likelihood can hardly be admitted. Even the little excavation that has taken place has revealed facts that are inconsistent wit htheora f imitationyo e workmanshiTh . o thoroughto s i pe Th . shape of the deeper type of ditch, the care bestowed upon its formation, and in particular the elaborate strengthening of its sides with wrought clay, all bear witness to the activity of experienced military engineers. Raedykes was no amateur improvisation. But for its marked irregu- larity of outline, few would have had much hesitation in accepting it as Roman. It therefore becomes important to inquire what weight should be attached to the objection just indicated. Attention has already been drawn to the fact that the six gates can be regarde s constitutina d g three pair e con f oppositesth o s - o t d an , sequent possibility of seeing in Raedykes a Roman camp of ordinary form, the sides of which have been deflected from the normal as the resulefforn a f accommodato t o tt e themselves e sinuositieth o t e th f so ground suggestioe Th . n will hardly appear extravagan thoso t to ewh familiae ar r wit e frui hth f recen o t t investigations reveales a , e th n di reports of the Limeskommission or in Professor Ritterling's masterly accoun e earlth f yo t Roman cam t Hofheipa e Tauiiusth n mi . These prove clearly that current notions as to the rigidity of the principles of ' castrametation' are only partially founded on fact.2 The rules were more elastic than antiquaries have been dispose f courseo , allowo t dis ,t I . beyond question that the typical Roman fort or camp was approximately rectangular with rounded corners, and that the great majority of known examples conform mor r less eequallo i s t i strictl e typet yth Bu o .yt beyond question that, when in the field, commanding officers felt themselve t a liberts o discart y e precepte th text-bookdth f o s f i , particular circumstances seemed to them to render drastic modification desirable. This is especially true of the days of the Republic and of the early Empire. And of all varieties of modifying circumstances those connected with the character of the terrain were naturally the most 1 Proceedings, xl. (1905-6) . 432p , . 2 See Ritterling, Das fruhromische Layer bei Hofheim im 1'aunus, p. 4. 346 PROCEEDING E SOCIETY , 19168 TH Y F .O S MA , compelling. Two or three concrete illustrations should suffice to show that the irregular outline of Raedykes need not prevent us from recognisin Romans a t gi . In 1898 there was discovered at Heldenbergen in the Wetterau, some little distance north-east of Frankfort-on-the-Main, the ditch of a large Roman camp of first-century date, which had stood on a bluff above a knee-like rive e benth f rdo Nidder. exaco N 1 t comparison with Raedykes

HELDENBERGEN

is possible. The outline had to be determined by the aid of sections cut at intervals, all surface traces having vanished completely, and the excavators were not fortunate enough to ascertain the whereabouts of any of the entrances. But the plan, so far as it was recoverable, takes n irregulae a for th f mo r polygo d bearn an (figo recognisabl , n s 7) . e resemblance whatever to the typical Roman entrenchment of writers on ' castrametation.' That it represents the handiwork of Roman soldiers

is non e leseth s indubitable evidence Th . e of finds conclusivewa s . ObergermaniscJie-Raetischer De e Se 1 Limes Bomerreiches,s de (Lief5 2 . . xiii.)Nr . THE ROMAN CAMP AT RAEDYKES. 347 Again t a Hofhei,e Taunusth n i m Professor Ritterlins hi d an g colleagues, after ten or twelve seasons of patient investigation, have been abl unraveo et histore first-centuro th l tw f yo y Roman forts that withie on e other. y nth la forn I 1 m (fig the) 8 . y were much more nearly oval than rectangular. The outer one, only a single entrance of which has so far been located, apparently belongs to the Flavian period. The inner, our knowledge of which is much more detailed, was probably constructe e reigs earlda th f Caligulas no ya . e Thregateth f seo have been identified with certaint, whilC) ed B, stronan , y(A g indicationf o s e positioth fourta f no h have come to light at D. It will be seen that these fall into pair f oppositeo s n i s much the same way as do the six at Raedykes t shoulI . addee db d that, as at Raedykes, the configuration of the ground supplies a ready explana- e deviationth l al tio f no s froe mth normal. Finallye fort th a te liff o th e, Waldmossiiigen in Wiirttemberg has been shown by the Limeskommission falo t l into stageso tw e lasf th ,o t earliee th n whici r d en h camn a o et half of the second century.2 This Fig.! throw firse th s t stage bac t leaska t as far as the Flavian period, when the fort was of earth. Even when subsequentls wa t i y rebuil f strangelo s stonen i t wa t yi , irregular shape. s originaIt l outline, however mucs wa , h more remarkable (fig. 9). Being

a permanent station, garrisoned a detachmenonl y b y f troopso t t i , 3 was naturally very much smaller than Raedykes, which was meant to hold an army. Otherwise, there is a curious resemblance, extending even to the manner in which the north-east corner is flung forward int e openoth . This resemblanc s nonei lese th es significant becaust ei is purely fortuitous, the character of the terrain being in both cases the determining factor. Fortifie thesy b d e Continental analogies e numbeth , f whico r h could readily be added to, we need not hesitate to set aside the main objection recognitioe th o t f Raedykeno Romans a s . And, once thas beeha t n dis- posed of, the conclusion seems fairly obvious, if due weight be attached to 1 See Ritterling, op, cit. 2 See Der Obe-rgermanische-Baetische Limes des Bomerreiches, Nr. 61b (Lief. vi.). earte gatee th Th h3n i s for t Waldmossingea t n hav t beeeno n satisfactorilyo tracedn t Bu . doubt there were four of them, just as there were in the stone fort that succeeded it. 348 PROCEEDINGS OF THE SOCIETY, MAY 8, 1916. the considerations that have been already adduced.1 Beyond the general statement, however, it is not in the meantime possible to go. We have heard that Roy assigned the camp to Agricola; Chalmers was equally confident that its builder was Lollius Urbicus; and others have preferred e wor th o regart f Severuss ko a t di . Eac f thesho e date s possiblei s . But there is no material on which to base a decision between them It is true thachampioe on t f Severuno s appeale ha se irregularit th o dt f yo shap s constitutinea gpresumptioa favoun s case.ni hi f 2o r But f ,thii s

WALDMOSSINGEN

Fig. particular featur admittes i e e s evidencda th n o onln t e allca ea yb t i , sid Agricolaf eo , sinc foue th e r f ever o parallel e yon s that were cited from abroad belonge e firsth t o dcenturyt . Ther muse ew contene b t t le o t t the problem rest. Some day, light from an unexpected quarter may show us where'-to'find the key.

. GLENMAILEKII . The site of this camp, now divided between the farms of Bush and Logie Newton, lies on the fringe of the uplands of Strathbogie, about midway betweea n s i Huntle easte wesFyvid t th th I an .tn n o eyo broad elevatiow lo , n bounded alon whole northers it gth f eo n aspecy b t a deep natural hollow, in the green bottom -of which run the infant supra.5 34 . p e Se 1 * Colone passage l Shanth e se ed: quoted infra.0 35 . p , THE ROMAN CAMP AT GLENMAILEN. 349 water Ythane th f hamlee so th ; Ythaf o t n Wells, named fro springe mth s in whics source it e rive s th h onls ha i r , hala ymila d f ean away s It . southern slopes command a wide sweep of open country. To the west, however, it is sheltered by a high ridge culminating in the flat rocky top

of Tillymorgan1 , over the shoulder of which peeps the distant peak of Bennachie. The ground between the foot of this ridge and the camp, thougtillagen i w markehs no ,i eighteenth-centurn do y map morassa s sa . Fro highese mth t e enclosurpointh f o t descene eth t toward Ythae sth n is fairly rapid, while beyon e streamdth , abov e farmhouseth f Grleno e - mailen or Glenmellan, rise two formidable hills with an open glen between. The strategic significance of the fortified lines is unmistakable. They were constructe armn a y ydb whic advanced hha d fro southe mth , whicd stild an reckoho ha lt n wit hhostila e force that might sweep down on it from north or east, and they are so laid out that on these two sides rivee th r wit marshs hit y banks gives effectual protection. The camp was first observed and surveyed, during the years 1785 and 1786, by Colonel Alexander Shand of Templeland, then a captain in the artillery, who at a later date returned to settle in his native Aberdeen- shire after a prolonged and strenuous spell of active service. Joining e rankprivatea th s a s wene h , t throug Sevee hth n Years' War, being severely wounde t Korbachda . Subsequentl American i s wa e ,y h wher e he was again wounded at Erandywine River, and finally he distinguished himself highly during the great siege of Gibraltar between 1780 and 1782. A zealous antiquar wellpracticas a y a s a l soldierkeenls wa e y,h interested in Roman roads and camps, and devoted much time to their investigation. In 1788 he prepared an account of his researches, including a description of Glenmailen, for the Literary and Antiquarian Society of Perth, which beed ha n founded four years earlier. Contrar expectationss hi o yt s hi , pape s nevewa r r printed e d firsonlth ,an ty volume Society'th f o e s Transaction t beinno s g published till nearly forty year d elapseha s d e interva(1827)th n manuscripI e . lth t seem havo st e been las e losttTh . we hear of it is in May 1788, when General Melvill, to whom as a leading authorit e subjecd beeth ha nn t yo i tsen r perusalfo t , writes Shanda complimentary letter tellthatd aboum an s requestedshi a , it ts ha e h , forwarded it to "the Rev? M* Whitaker"—that is, no doubt, John Whitaker e well-knowth , n historia f Manchester.no e absencth n I f 1eo this authentic record of his activities, we may fall back on a passage in Newte's Tour which Shand himself expressly authorises us to accept as accurate, stipulating only thashoro tw tomitted e t b phrase o t t I e . sar wil seee lb n that these suggest some measur f indebtedneseo Melvillo t s , 1 Melvill's letter to Shand is reprinted in Proceedings, vii. (1866-67), pp. 29 ff. It has a special interest as containin ghriea f accoun Melvill'f o t discoveriesn sow . 0 35 PROCEEDING E SOCIETY , 19168 TH Y F O S. MA ,

which Shand obviously regarde originalityn reflectins da ow s e hi Th n g.o repudiation, however, is made with every mark of old-world courtesy: " the editor having been misled in that assertion concerning an Officer, eminent for his critical knowledge of the Roman classics and Roman British topography." followinthe In 1 g extrac wordthe t whicto s h Shand took exceptio placee nar squarn di e brackets. Referrin Strathmoreo gt , Newte says:2— "The chai f Romano n camps in this great Strat firss hwa t discovered, as already mentioned Lieutenaiit-Generay b , l Melvill 1754n afterd i , an , - wards very accurately delineate d describean d Major-Generay db l Roy. Captain Shand, [from the example, and at the instigation of General Melvill], embraced the opportunity which a four years' residence at Perth, with the command of the Royal Artillery, in North Rritain, afforded him, of exploring the Roman geography in Scotland, and comparing the Roman field-works and engineering with what he had seen practised in the German and American wars. By a narrow inspection of Strathallan, Stratherne Strathmored ,an t onlno ye traced,h , [after General Melvill]e ,th great consular road, with the numerous posts, prcesidia and castellci, as wel greas la t camps situate near o , from-Cameloin rdit o Kerrymuio t i n i r Angus e ceasediscernibleA b , , wherVI o afterwardst s he e th t e bu : , discovere verda y great numbe f vicinaro r croso l s roads nea rivere rth s Erne and Tay, and visited the other Roman posts as far as the end of the Grampian Hills near Stonehaven, whic beed hha n suppose havo dt e been e remotesth tcountre pointh f o twhico yt e Romanshth laudy b d , ha , penetrated. Conceiving this opinio inconsistene b o nt t wit warlike hth e character, and mighty exertions of the Romans, lie likewise, in search of Roman antiquities, explore e Countriedth North-Grampianse th f so d an , foun e greath d t cam t Glen-Mailepa Ythann o n , perhap e statioth s d a Ithwnam, the very remarkable presidium near Old Meldrum, with a numbe smallef ro r work l similasal othee thoso rt th n reo side havind an , g e samth e kin f charactedo r; onl yf the o som t executemw no fe e d with such nice accuracy: a circumstance which may be owing, perhaps, to their being a century or two later than those of Agricola." e earliesThith s i s t referenc Glenmaileo t e n which appeare printn di . Apart fro explicie mth t imprimatur already allude , therdto abundans ei t evidence of an internal kind to prove that the whole passage was inspired by Shand knowe W r instance. fo , , from other sourcet no d s di tha e h t certainte th Strathmore o shart th s viewe f a y o eth y f Melvilso Ro ed an l camps being Agricolan. Again, the allusion to "the very remarkable presidium nea Meldrud rOl m " agrees closely with Shand' description sow n o f Castellue "th Barra-hilln mo Meldrumd , nigOl . hto y statioa , wa o nn inferio grandeurn i r goor o , d preservation e kindwory th an f k,o o that , t t Ardoca h excepted."3 Incidentally explainee b y ma t d,i tha so-callee tth d

'Notee th e Shany b 'Se d reprinte Proceedings,n i d vii. (1866-67) ff.7 ,2 fro. mpp , which mosf to

the1 foregoing particulars have been drawn. 2 Thomas Newte TourA , Englandn i Scotlandd an . f 1 (1791)30 . pp , 3 See Proceedings, vii. (1866-67), p. 28. THE ROMAN CAMP AT GLENMAILEN. 351 " presidium " or " castellum " on the Barra Hill is a. circular fort surrounded by several concentric ditches; Chalmers is doubtless right in regarding it nativa s a same forth f e well-knowo te clas th s sa n Caterthuns. Finally,

the allusion to " the statio ad Ithunam1 " points clearly to the influence of studena s Richar' wa f o o ton eCirencester.wh f do ' Precisely the same features can be detected in the " short account of came th p near Glen-mailen r JohSi nn i , Sinclair's Statistical Historf yo Scotland, vol. 12, at pages 287, 288, & 313 to 316," which Shand likewise commend e attentioth o st e curious, f "nth o " pendin e publicatiogth f no his own paper by the Perth Society.2 The parish ministers of Forgue d Auchterlessan noticeo tw e writere sth , th whic f o se citesh , were neighbours of Shand's, and what they have to say upon the subject obviously reflect conversations hi s lattee Th .r (the Rev. Alexander Rose) give t mucsi more hth e generous allowanc gathere spacef on eo d s an ,tha t his observations had been submitted to Shand for his criticism.3 He describes the camp in some detail, and adds a long footnote, the following extract from which may be compared with the passage quoted from Newte:— werstupendoue o authore th "el th Wh al f so s military works, whether road placer so f defenceso , scattered longeoveo n e country th re rar e w , losa knowo t t s a ingenioun A . worthd an s y gentleman nativa ,e th f eo neighbouring parish of Forgue, and who has served as an officer in the Royal corps of Artillery, since the year 1758, was desirous to compare seed whaha n time e n durinth ow , s witghi h what coul stile db l tracen di countrye th f Romao , n field fortification othed an , r topographical marks of their wise military institutions. His situation at Perth, in the duties s professionohi f 1787f , o fro d e ,yea mgavth en oppor rm e ehi th 178 o -t 5 tunities of spending a great deal of his spare time, in these wished-for researches, which having pursued with unremitting assiduity, he was at last enabled, contrary to an opinion which then prevailed, to demonstrate that the Roman armies had passed the Grampians by land, as well as that they had surrounded the cost-land by their snipping ; the character, style, manned an f fielro d fortification, bein evidents g a wels a d l supportean , d e Castelluinth e CastrBarra-hilln mth o n i a sestivd an , t Glenmailena , as any where between them, and the prsetentura of Agricola extending from Fort Clyde."o ht « Another extract froe sammth e footnote, wit s referencit h o t e e Militaryth Antiquities, will serv o introduct e w poinne f o ta e interest:— "The ingenious autho e investigationsth rf o (Captai ) S. . A n, just recitedt firsa s t inclinewa , believo dt e Statieth Ithunamd oa e th s wa ,

Caledonia, vol. i. p. 90, with a plan. Proceedings, I.e. 2

lase tTh 1 paragrap lona f hgo footnote begins same : "Th e ingenious gentleman whoo t , e mth publi3 s indebteci r thesdfo e observations, takes notice that something more should have been introduced about the Roman roads," etc., etc. 4 Sinclair's Statistical Account of Scotland, xii. p. 314. 352 PROCEEDING E SOCIETY , 19168 TH Y F .O S MA , work of the Emperor Severus, yet some of the best informed, and learned antiquaries will have it, that all the posts N. of the Grampians were constructed by Lollius Urbicus, the brave and gallant Lieutenant of Antoninus; and the late ingenious Major-General Roy, as soon as he peruse e pla dth f Glenmailei no environsd an i , wit s explanatiohit d nan references, put it down immediately in his Mappa Britannise Septeii- trionalis, Castra Agricolse. Therefore, it is to be presumed, it will be piiblished in the next edition of the General's map of Scotland,1 and that some account of it will be given in the Appendix to his Posthumous Work probablw no , ySociete printinth y b f Londo yo f gof n Antiquaries, manuscripts hi whof o t o e mbequeathes on swa lattey db r will othee th , r remainin King'e th n gi s library." " Vol. xii. of the Statistical Account was published in 1794. The foot- note just quoted must, however, have been penne t leasda yeaa t r earlier, for Roy's posthumous work was issued by the Society of Antiquaries in 1793. The writer's expectation was only partially realised. On the " Mappa Britanniae Septentrionalis," which forms plate I. of the Military Antiquities, e positioth f Glenmaileno s duli n y marked nea e headth r - waterYthane th f so , while plat . containeLI e enclosure splaa th f no s a , well as a section of the rampart and ditch. On the other hand, not only

is there no "account of it" given, but it i3 s not alluded to in the most distant way either in the text or in the appendixes. The analogy with Raedykes is thus very close. And the parallel extends further. On both plates the Glenmailen camp is actually designated " Re-dykes." It would be interesting to know whether that name was ever really applied to it, r whethe o facee ar de w rwit e resulhth f somo t e confusion trus i t eI . that Chalmers asserts tha tcame "th t Glenmailenpa came wels th a , s pla at Urie, is called the Rae-dykes, from the Gaelic Ra, signifying a cleared spot, a fortress."4 But this maybe based on no better authority than the plates in the Military Antiquities. What is certain is that to those eighteenth-century writer kneo e districwswh th t firsa t t hand—Shand o paristw e hhimsel th minister d e Statisticalan th f f o s Account—e th enclosure was simply " the camp at Glenmailen." Mr Rose, indeed, in his footnote speaks of "the appellations of ri-dikes, and grim-dikes" in a way that suggests that they were used by the country people in con- nection, not with the camp, but with other earthworks in the district.

Possibl nomenclatur cluthe ythe e to e adopte thdin e Military Antiquities 5 e proximitth lie n e Ri-hilsi th f r Re-hillyo o l , for t whicA fige . hse .10 l eventsal , that some distrus s justifiabli t t oi f cleas ei r from what Shand

"e firs1Th t edition Boy' f wa"p o ssma engrave 177n di 4 (Gough's British Topography (1780). ii , 561 and 586). The only copy I know is in the British Museum. Sinclair's Statistical Account, . 315p . relevane Th 32 t portio e platth s herf ei no e reproduce scala n o e , figsomewhas da 10 . t smaller tha originale nth . 4 Caledonia, vol . 127p . .,i footnot • e (g). 5 Statistical Account, xii . 313.p . Section of the Va//i/m oDitrA (tt

20 lf_ _ 1C 5 o Scale offrst.

PLAN

of Grounds Parishesthe tn of Forg, Achterlefs, and Culsamon ; arifZ COUNTY or ABERDEEN , . ej~7lioitin(/ the dncientftim/j o/ jLogt? •? ' -' Wtivton near Glen mat fen,an the So/ith Bank ef ',-1 Ythdn, tof/etTitr icith thf Mifitnrv //V/<*.i ' a rid JntrfiichrtientA' n^ar Tif/i'ttrrrf/n.

REFERENCE.

A.B.C.D.E.F./V?'- Camp of Re-,Jykes. the ftt-

ripJirrv of which is 3140 Y'ar>fs,-near '2 Statute CO nd cotnjirelitndint? s^n>ftftifi(j wore than

00 Scotch Arrex,

Fig. 10. Plan of Glcnmailen, from Roy's Military Antiquities. (Scale: ^.tair (roughly).) 354 PROCEEDING E SOCIETY , 191u8 TH Y F .O S MA , e 'Noteth say n i s' whic s alreadha h y been quoted more than once. e thuH s expresses himself regardin e reproductioth g s plahi n i nf o n Roy's book:— " Captain S. did not finish his plan of the ancient military vestiges, on the sources of the Ythan, till some time afterwards, and permitted a good many copies to be taken, and as some of these have been copied from other copieserrorw fe sa , have crep , particularltin orthographye th n yi , several of the names of places, and grounds in plate 51, being spelled in suc ha manne r as would make them unintelligibl e countrth o t e y inhabitants."1 This extrac s interestini t g from another poin f viewphrase o t Th . e " some time afterwards " refers back to " during the years 1785 and 1786," the period when the original discovery and survey of Glenmailen were made e completTh . e plan was, therefore, probably preparee th r fo d communication made to the Perth Society in 1788 and submitted to General Melvill in the spring of the same year.2 Very possibly Melvill may have been the intermediary through whom the plan reached Hoy.3 In 1788 the latter was deeply absorbed in the preparation of his final report on the great trigonometrical survey which was to determine the relative position observatore th f so t Greenwicya f thao t Parisa td han . This occupied all the energy that ill-health allowed him till his sudden deat Juln hi y pla e 1790 Glenmailef nTh o . lais dnwa asid anticipation ei n otima f f leisureo t destineeno thas arrivetwa o d t . Apparentls wa t yi not even copied, as that of Raedykes had been; No. LI. is the only one of platee th whicr sfo h theroriginao n s ei l extant cop. f yo MS , eithee th n ri Militarythe Antiquities preserve Britisthe in dh Museu thain tmor belongin Societe th o g t f Antiquariesyo ownee drawine th Th f . ro g from whic e engravehth r worked—whether Shand himsel r another—muso f t have reclaime t aftedi e boopublisheds th r kwa . t wilI rememberee b l d that, accordin Reve Roser th M . o gt ,late "th e ingenious Major-General Roy e peruses sooh a , s a ne pla th df Glen no - mailen and environs, put it down immediately in his Mappa Britannise Septentrionalis, Castra Agricolae." The statement is not literally true. Roy did not insert the camp in his map at all; that was done by his editors. Nor is there, as there was in the case of Raedykes, any direct e hanth evidenc f Agricolat di o sam e n i th t eew A timthasa . e eth there

1 Proceedings, addee b y viid ma . tha t (1866-67)I t. recen28 . p , t inquirie localite th n si y have failed to bring to light any evidence that the camp is ever called Re-dykes by the people of the district. supra.9 34 . p e Se 2 3 This possibility is in no way inconsistent with the statement of Chalmers (Caledonia, vol. i. p. 127, footnote (g)) to the effect that Shand "communicated his discovery first to the antiquarian society at Perth in 1788, and afterwards his survey of it to General Roy." E ROMATH N CAM T GLENMAILEN5 A P 35 . is no reasonable doubt as to what his view would have been, had he found occasion to express it in writing, for Glenmailen presents all the characteristic features which led him to assign other temporary camps to the Agricolan rather than to the Antonine epoch. On the other hand, s editorhi s were more cautious. Whil t hesitate no the d ydescribdi o et e Raedykes as " Cast. Agricolce" they were content to attach to Glenmailen —or Re-dykes s the,a y called it—the more general descriptio f "o n Cast. Rom." Either they felt it incumbent on them to be careful, in the y absencdefinitan f o e authority from Roy r the,o y consideret i d impoliti o disregart c e opinioth d s incline e f Shandwa nw o s o (a d wh , know fro Rose'r mM s footnote attributo t ) e Glenmaile Severuso nt . Thus muc r Roy'fo h s mape questioTh . f origio n s handlei n d even more circumspectl e explanatorth n i y y notes which occup e margith y f o n plate LI.d whican , h doubtless derive ultimately from Shand himself. There the only allusion to the matter is in the description of the three surviving gates, whic "e coveresaie b har o d t d with Traverse same th en si manner as the Camps which are supposed Roman, on the South side of the Grampian Hills." It would not be easy to imagine anything more definitely non-committal. Coinin enclosure th o gt e itself shale ,w l find referencn o , fig thao , et .10 t 178n i 6 enoug e fortificationth f ho s remaine givo dt ecleaa e r ideth f ao origina l\vholee for came th beef md Th o p.ha n lozenge-shape outlinen di , and the motive for this departure from the ordinary rectangular design is self-evident. By planting the northern corner within a bend of the Ythan makind an , interioe gth r angle decidedly obtuse t becam,i e possible to secure the river as a protection along the full extent of two sides. For the greater part of this distance—that is, along the north-western front —the defences ran close to the brow of the natural hollow which the streabeddistancits The carve.as m has eout d fro edgthe m the eof escarpment is seldom more than 40 or 50 paces. As the water was from 50 to 80 feet below, the obstacle was most formidable. Of the gates, which were apparently six in number, we have already heard. Three of them were visible in Shand's day, together with the traverses that had stood before them dimensionThe . sidethe s of s canno estimatebe t d wit accuracyy han smallo e plaexplanatorn to e scal th A s th :ni f . eo y note, however, states that "the periphery ... is 8140 Yards, near 2 Statute Miles." As the proportion of length to breadth is about 3 :2, we shal fairle lb y saf n assumini e n arega f 282ao 0 fee 189x t 0 feet r abouo ; t 122 acres, which agrees reasonably well with the " something more than 90 Scotch acrese explanatorth f "o y note.1 Later writers make but brief reference to Glenmailen, and from none 1 The Scotch acre contained 6084 square yards. S56 PROCEEDING E SOCIETY , 19168 TH Y F O S. MA , t fresoge f he the w o informatio d mo face e d on w tvalue y e an Th f .no learn is that the process of destruction made rapid headway during the half century that succeeded the publication of the Military Antiquities. Thus in 1845 the parish minister reports that " the south and west dikes onl entire."e yar 1 Interprete lighe th f fig thi, o ty db .10 s must mean that south-ease onlth n yo south-wesd tan t sides were remain rampare th f so t to be seen in anything like their original condition. It cannot mean that e south-easth t s completesidwa e r nearlfo , y three-fourthd ha t i f o s disappeared before Shand's survey was carried out. It implies, however, that the whole of the north-west side, which was in relatively good preservation in 1786, had been ploughed down in the interval. In short, the state of matters in 1845 would seem to have been very much the same as that which prevails to-day, when (as indicated on the Ordnance Survey Map e rampar)th s visibl i ta continuou r fo e s stretc f abouho t e easterth fee0 t a 86 t n extremit •e south-easth f yo t side d agaia an ,r nfo similar stretc e westerf abouhth o fee 0 t a t95 t n extremit southe th f yo - west side e curvbeind th , en eac n ege cornei e oth th hf t casa r e clearly discernible. Further, along certain portions of the north-west and north- east s sidestili t li s possibl o recognist e e defences th e lin f th eo e f i , e conditioth e cropth f sno be o admit suca thorough-goin s f a ho t g scrutiny.2 ...•,... The present-day surface-appearances, then, are entirely consistent . wit e pla hth s lai na d dow e samShandy th b n t e A tim. e these ar e admittedly fragmentary t therefori d an , e seemed desirabl probo et e eth matter further, before accepting Roy's plate LI. as a definite basis for argument. This was the motive immediately underlying the little expedition organised by Professor Haverfield and myself in July 1913. Leave to excavate was readily granted by the proprietor, Mr Garden Duff of Hatton, whil r AlexandeM e r Hay ,e tenanfarth f Bushmo f o t , extended a cordial welcome to the excavators, and saved them both tim troubld ean recruitiny eb gbana intelligenf do unusualld an t y hard- working labourers e .exigencie th Eve e cropd th nha f so s allowede th , brief week at our disposal was not long enough to justify any such attemp systematia t a t c exploratio subsequentle s outlinth wa f ns o e a y attempte t Raedykesda t wasi s e fieldmanA o ,s .th f se yo werth r efo 1 New Statistical Account, vol. xii. (Aberdeen), p. 286. 2 For the information as to the north-west and north-east sides I am indebted to Mr J. Graham Callander, Secretar e Societyth f r Callanderyo M . , whose experienc n connectioi e n wite hth Ancient Monuments Commission lends great weight to his opinion, was good enough to examine e grounth d very carefull behaly m Octoben i n fyo r 1916, afte crope beeth d r sha n cleared away. When Professor Haverfield and I visited the site, access to the fields concerned was unfortunately t practicableno Callander M . r tell intere sm alia thamarkinge th t s describe "Earthworks da " O.Se th f . o 25-in (1901)4 p in68 ma .fiel. , whatevedNo connectere b the representy o t yt ma no d e ar , with the north-west side of the camp. THE ROMAN CAMP AT GLENMAILEN. 357

moment inaccessible that attentio s perforcnwa e concentratee th n do ground lying towards the western end of the enclosure. Cutting s e northermadth t ea n extremit e south-westerth f yo n face revealed the character of ditch and rampart. The former, which had had sloping sides, must have bee fee 9 t leasn a r t tleaso a wid 8 t d feet4 ean t deep. The latter, which must originally have been about 20 feet thick, was constructe loosf do e eart stonesd han ,t seemed i lai s dspeciall(a a n )o y prepare f whito d r yelloeo dbe w clay ,inche3 somr o thicknessn i s2 e . s outeIt r s practicallfacwa e y continuous wit e scarhth r innepo r face ditche oth f ; that Raedykes, jusa is ts a t , ther beed eha n nothine th n gi nature of a 'berm.' The whole of these features, except of course the underlyin f clay o e reflecte d ar , gbe d with tolerable clearnes Shand'n si s section, which is shown in the upper right-hand corner of fig. 10. This, t wil i seen e takes b l wa , nmose closth to et northerl y whol e pointh f eo t camp, presumably because in 1786 it was there that the conditions for obtainin a grecor d were most favourable. To-daye knoww s a ,, that particular portio defencee th f ns o bee thoroughlsha o ns e yb razeo t s da barely discernible. Towards the eastern extremity of the south-eastern side, however, the rampart still stands some 5 to 5|- feet high on the insid d somo 7t £an e 7 efee e toutside th hig n ho , whil e greatesth e t apparent breadt base th s abouei t ha feet8 1 t . Elsewher s stateit f eo preservation is less satisfactory. Along the south-west face, for instance, e maximuth m height attainee outsidth fee5 aboud n o 4s o tdean i t J t a foo e insidet th les feen 2 o s1 , twhil o t represent 1 e1 e maximuth s m width at the base.1 The discovery thaf yellowisrampare o d tth be rested a ha ht n clado y furnishe guidda f materiaeo thas twa l assistanc tracinn ei originae gth l defencee linth f eo s over ground wher outwaro en d mar f theiko r former presence remained visible . neas i Thid r senougbe surface e th b o ht o t e easily reached wit spadee hth yetd an ,; having escaped serious disturb- ance whe e superstructurnth s destroyedewa sufficientls i t ,i r dowfa y n immune b o t e fro ravagee mploughe s th aid it th ,y f coupleo sB . d with occasionan a l clearing-ou ditce th front n hi f o te souther th , n corned ran the now vanished rampart at F were shown to have been correctly

laid down on the plan in the Military2 Antiquities. The site of gate A wit s traversit h s unfortunatelewa y covered with growing corn; but, by working along the north-western face from the western corner, the position of gate B was determined and found to agree with that assigned to it by Shand. The gap by which it was located was 70 feet e normath wide t A l. distance befor , digginit e g brough o light t t e traversee ditcth th f o h , immediatel e e rea th f th whic o rs n yi wa h 1 I am indebted to Mr Callander for verifying these particulars. * See fig. 10. 358 PROCEEDINGS OF THE SOCIETY, MAY 8, 1916. bed of yellow clay that had formed a foundation for the protecting mound worda n I ,. whereve s accuracit r testeds ywa , Roy's illustration stood the ordeal in a manner which justifies us in accepting it as a reliable representatio e camt appearei th s f pa no 1786n di . This gives s threu e gates with traverses frod an m, these s wil a ,e eviden b l t from quity a glance planma e th x safeltotaa t ,si ea f inferrede o lb y . The clearing of the very first section of ditch and rampart left on excavatore e mindth th f o s distinca s t impr'ession that they were face faco t e wit e worhth f Roma ko n wore engineersth ks a proceeded an , d the impression rapidly became a certainty. The form and design of e wholeth e arrangementh , e gates e th ditch e slopth f e th o ,f t th ,o e extensive vise of clay, the presence of the traverses—all these are characteristic. The camp at Glenmailen may confidently be set down a memoria s e a largesth f o lt Roman army that ever penetrateo t d the remoter portions of our island. It must, of course, have been a fiel d perioe s stath forceit yf d d o woul an , d necessaril briefe yb . Hence the absence of relics. The ditch was thoroughly examined at several points, notably near gate B and at the adjacent corner, and the black matter from the bottom carefully scanned; but no coin, no fragment of pottery, emerged to reward the diligence of the searchers by giving a singl e themfindf s mado th e wa sp f a databl u eo m su e e clueTh . piece of iron, which was lying fairly high up, and which may or may not be Roman, a few broken twigs of ancient forest trees, and two or three masses of rust which possibly represented corroded spear-heads. Exploratory trenches cut as near the centre of the camp as was practicable proved equally unfruitful. Tha t reveat no y the d an ly di trac f buildingo e s scarceli s y surprising. Officer d soldieran s s alike woul e quartereb d e thuar e n tentssi th e d lefn doubW o i t .t s a t precise epoc o whict he entrenchmenth h t belongs. t a Here s a , Raedykes muse w , t wai r furthefo t r presenlighte Th . t positios ha n been excellently stated by Professor Haverfield:—* "The general result is a plan which is a Roman plan and which includes at least one gate in a style used only by the Romans. One may therefore welcome the encampment as Roman, and may further deduce, from the absence of small finds and from the exceedingly wid e rathee th gate d r an sslende r defences, thae occupatioth t s nwa very short t i was; n ,facti e 'marchinth , g n camparma f f o yo ' 10,00 r 15,000o 0 men, abidin nights w r onlfe gfo ya . Some day, further digging may tell us the date of the work within the Roman period.

1 "Roman Britain in 1913" (British Academy Supplemental Papers, ii.), pp. 8 f. In the fore- goin gI hav e draw somo nt e extent upon this report Professoy b r Haverfield, and, still more freely, upon an article of my own, published in the Scotsman of August 2, 1913. THE ROMAN CAMP AT G-LENMAILEN. 359

The choice mus e betweeli t n Agricola abou t d SeptimiuA.Dan 3 .8 s Severus about A.D. 209 y seeA priori,. ma m t i likel y enough that Ythanort the far tha Agricol and nas has t Severugot a notdid s , but only actual finds can decide. I will, for the present, add one warning. The site of our camp agrees ill with the description of the in the ' Agricola' of Tacitus, except on the assumption thae Romanth t s y thafaceda t d south d thae an th ,t Ythan guarded their rear. No battle in the least resembling that sketche y Tacitub d s could have been fought here wite Romath h n front restinriver.e th n g"o POSTSCRIPT. On p. 340 supra, reference is made to the surprisingly accurate mannee mosth f t n whico i obscurr e on h e section e outlinth f f o so e Raedykes has been reconstructed on the Ordnance Survey Map. Through the courtesy of the Director-General I have now been able to consult e 'Name-bookth e paristh f r Fetteressoo hfo ' . This shows thae th t surve mads y wa Augus n ei t 1864 thad relativs ,an tit e accurac entirels yi y due to the great care and thoroughness with which the responsible officer (Corporal Render, R.B.) performed his task. He reports that " General Roy's plan is pretty good on the North, East, and West sides, but very muc Soute erron hth i n hro side, thatd "an , where both rampart and ditch had disappeared, his own conclusions were based on a close e undulationstudth f yo f tho s e ground, fortifie "thy b d e assistancf eo the adjoining tenants who levelled those portions during the process of cultivation. n regarpositioe I "th o d t f gat no , witeE s traversehit , simple h y follows Roy r ratheo , r Brown.