EXCAVATION T BIRRENSA S , 1936-1937. 275

I.

EXCAVATION BIRRENST SA , 1936-1937 ERIY B C. BIRLEY, M.A., F.S.A.ScoT., F.S.A., WITH CONTRIBUTION . RICHMONDA . I Y B S , M.A., F.S.A.ScoT., F.S.A., J. A. STANFIELD, AND W. PERCY HEDLEY, F.S.A.

presentinn I g thi allowese b report y expreso dt ma 1I thank y sm o t s collaboratorsy m . RichmonA . I r M d. pai dvisita , which proved most instructive e excavationth o t , f 1936o s , sharee directioth e n th i d f o n work undertaken in the following season, and has contributed the sections of this report which describ s waterit e rampar e ford th e th -an tf o t supply, besides giving general help wit sectione ht th no whicr s i sfo e hh directly responsible r RichmonM ; d undertoo e surveyinth ke th f o g

1 The following abbreviations are employed:— AAa~4 . . . Archceologia JEliana, 2nd—4th series. CIL . . . Corpus Inscriptionum Latinarum. . Transactions, . e Cumberlanseriesth CWw . f ne ao , d Westmorlanan d d Antiquarian and Archaeological Society. . Folze . r B. P81zer Bilderschusselne Di , der ostgallischen Sigillata-Manufakturen, Bonn, 1913. Hofheim . . E. Bitterling, Das fruhromische Lager bei Hofheim i. T., Wiesbaden, 1913. JRS . . . Journal of Roman Studies. . NiederbieberOelmannF . Keramike Di , . Acs Kastells Niederbieber, Frankfurt a.M., 1914. NMA . . . National Museum of Antiquities. . OswaldF . , . Index f Figure-typeso . n Terra. o 0 Sigillata, Liverpool, 1936-1937. obergermanisch-ratischer De . . OB . L Limes Romerreichess de (Reporte th f o s German Limes Commission). PSAS . . . Proceedings of the Society of Antiquaries of Scotland. Bicken, Saalburg . H . Bicken Bilderschusselne Di , Kasteller de Saalburg Zugmantel,d un in Saalburg Jahrbuch, viii., 1934. r GeorgSi . e Macdonald . BWS Romane . 2Th , Wall Scotland,n i second edition, 1934. Well-known excavation report e indicatee sit ar se namn italics;th i eth f o ey t b di wil e b l convenien givo t t e further references her thoso t e e sites which have been reporter o d A upoA n ni CW: . AACorbridge,3, . viii. 1911. AA4, vii. . . Denton Hall turret, Chapel House milecastle. . Corbridge, . AA . 19384xv , . CW2, xi. . . Poltross Burn milecastle. CW. High2, xiii. House. milecastle, Birdoswald, High House, Appletreed an turrets, Throp fort. CW . Birdoswald1, xxx . . fort. VOL. LXXII. 19 276 PROCEEDINGS OF THE SOCIETY, MAY 9, 1938.

structures examine s alsha o d draw1937n di an e figur, nth e illustrating .the two west gateways, found in 1895 and 1936 respectively. Mr J. A. Stanfield, besides drawing most of the decorated Samian ware and the bronzes, has contributed valuable notes on the most interesting of the former material; and Mr W. Percy Hedley has provided notes on the coins. During both seasons Mr W. L. George, B.A., assisted in the direction excavationse ogiveth s f ha ne h considerabld an , eexaminatioe helth n pi n of the pottery; I have had the .benefit of consulting Mr F. G. Simpson, Hon. F.S.A.Scot. connection i , n with e lattesomth f ro e materialr M . . ReiC d. R undertoo businese kth s arrangementst thai e o th t w sa d an , detaile"wore b o kt d presently curtaileshoule b t dlacno r dfundf fo k o r o s labour; and Thomas Batey, my foreman, showed himself equally competent as an excavator and as a trainer of the local men whom we employed.

. INTRODUCTIONI . The original excavations at Birrens in 1895 1 aimed at proving the Roman date of the visible defences and securing as complete a plan as possibl interioe th f eo r arrangement e fortbotd th an f h; so objects were attained, wit h succesa mor e th l e sal noteworth y becaus stimulats it f eo - g influenc e in developmenth n o e f scientifio t c excavation elsewhere. Furthermore, considerable incidental light was thrown on the history of Birrens in Roman times, not merely by the inscriptions, coins, and other relic e discoverth s y foundb t f evidencyo bu , drastia r efo c rebuildinf go e forth t afte a violenr t destruction attempo n t s madBu wa o tt e. discover o structurawhethetw e th r l periods between them accounted whole foth r onlr eo occupatioe ypara th f o t Birrenf nRomanse o th y sb ; and, with t potterestablisheye t yno meana s a df dating o s , thers ewa no guide other than that provide inscriptione th y db e coind th an so t s time when the fort "was first built or to the total length of the occupa- tion. Among the inscriptions, only one was directly dated, to A.D. 158,2 nond an e coul show e earliee db b lateo r nt o r r tha secone nth d century; the brief coin list gave no cause to assume an occupation antecedent or subsequen thato t t century seemeit and follo dto ; w thahistorthe t y

of Birrens must have3 been simila y Antonin an o that rf o t e forn i t Scotland, established befor e middleth e d finallan , y abandoned before secone th f ende o dth , century. tims a t e went becamBu i n o t e clear that Birren splace e coulb t dno

1 PSAS, xxx., 1896, pp. 81-199; reprint, pp. 1-119. 2 Ephemeris Epigraphica, . 123ix 0 (correctin readine gth g given in PSAS, . 129)xxxp .. 3 Cf. PSAS, lii., . 2171918p ., EXCAVATION BIRRENST A S 7 , 27 1936-1937.

e samth en i categor s JSTewsteaya e Antonin e fortth th f o sr o d e Vallum. The historical attributio differene th f no t structural period t Newsteasa d can hardly be settled without further digging,1 and here we must confine ourselves to noting the mere fact that they outnumbered the two periods at Birrens. But excavation in the forts of the northern limes regularly produced three periods attributable to the occupation inaugurated by Lollius Urbicus, and it became clear that the discrepancy could only be explained in one or more of three ways: (1) Birrens had been occupied fo shortea r r period tha othee nth r forts prioria likela ; t yno s thiswa d succeedesuggestionha t e I destruction ) th n escapin i d(2 f o . e gon s which overwhelme forte dth s farther north; this possibilit s clearlywa y strengthened by the results, to be referred to later, obtained by excavatio t Risinghaa n d Higm an o period h tw Rochester e s Th ) (3 . note 189n di 5 represente dpara t s occupationonlit f yo I wil; l show presently that this explanation seemed certain to me, but I must first deal briefly wit hdifferena t readin e evidenceth f go , whic largels hwa y the undertakiny causm f eo g excavatio t Birrensna . e concludinth n I g chapte e seconth f o rd editio s Romanhi f no Wall in Scotland, Sir George Macdonald, adopting the second of the above explanations, mad s interpretatiohi e e evidencth f o n e from Birrena s touchstone for the vicissitudes of the Antonine frontier. His argument may be summarised thus: 2 The inscription of A.D. 158, already referred , musto o observetw t e belone seconth th d f o t go periodsde th r fo , circumstances unde r s founwhicwa d t i h showe s stild n i wa thal t i t position when Birrens was deserted for the last time; there was no break in the occupation of the site between that date and the final destructio archaeologicae th forte d th an ;f no l evidence showed thae th t sitlaid eha n derelict after circa A.D. 196. Ergo, Birrens (like Hadrian's Wall escaped ha ) disastee dth r which overwhelme Antonine dth e Vallum earle i nth t occupie yno years Commodusf wa dso t i late f i rd thaan ; n A.D. 196, the Antonine Itinerary, in which one route has Birrens as its northern terminus, must be assigned to the time of Commodus at latest; and the absence from it of the forts farther north showed that they, and the Antonine Vallum, must already have been abandoned under that emperor. place th discuso t et Thi no widee s si sth r histore questioth o t y s na e Romaoth f n Wal Scotlandn i l , thoug e sequehth l will show thay m t interpretatio evidence th f no e differs "widely fro r GteorgmSi e Macdonald's; s necessari t i t fort t se e reason th ho bu yt o reject se e whicth tm d hle

1 Cf. Dragendorff's acute observations in JRS, i., 1911, p. 135. 2 RWS", p. 478. 278 PROCEEDING E SOCIETY , 19389 TH Y F .O S MA ,

above interpretatio e evidencth f no e from Birrensinitiato t d e an eth , recent excavations ther orden ei provo t r point.y em 1 In the first place, I concurred in thinking that Birrens had escaped the disaster of circa A.D. 180, because its position suggested that Birrens should be regarded as an element in the Hadrianic frontier rather than e Roma th a lin n s i ka n occupatio f Scotlando n e accompanyinTh . g sketch-map (fig ) wil1 . l serv o emphasiset e closeth e similarity between s positioit thad Netherbf nan o t r Bewcastlyo relation ei Hadrian'o nt s

TO FORTH

HIGH* ROCHESTER

Fig. 1. Hadrian's Wall and its outliers.

Wall, and those forts are known by inscriptions 2 to have been estab- lished by Hadrian. The three forts form a screen, thrust forward eight or nine miles beyon e Walldth , int aren oa a where archaeology suggests that ther thes enwa considerabla e native population. Ther s beeeha n no scientific excavation at Netherby, and Bewcastle did not receive attention until 1937,3 but Mr Richmond's excavations at Risingham and High Rocheste 193n ri 5 4 showed that both fort escaped sha destruce dth - tion which overwhelme Antonine dth e tim e Wal th Commodusf eo n li d ;an if they escaped, it seemed unlikely that the nearer outliers in the west, less easily accessible from the northern limes, should have been affected. preliminar1A y outlin viewe th f eso her t fort givens e se revie a hwa n i BWSf wo Proceedingsn 2i Societye oth f of Antiquaries f Newcastle-upon-Tyne,o 1 (1934) 28 h series . 4t p ; . vi of., also Dumfries- shire and Galloway Transactions, xx. pp. 157-170, from which the following discussion has been expanded. 2 CIL, vii. 961, Netherby; 978, Bewcastle. 3 Cf. OW2, xxxviii. pp. 195-237. 4 AA4, xiii. pp. 170-198. EXCAVATIONS AT BIRRENS, 1936-1937. 279

But the evidence from High Rochester seemed even more apposite to the study of Birrens, because until 1895 no Roman site in the north of Britain had been so completely examined. As in the case of Birrens in 1895, High Rochester, in the middle of last century, was examined with a view to obtaining a plan of its interior arrangements, and the excavators distinguished distinco tracetw f o st period f occupationo s , which could be matched with building-inscriptions of two periods, Antoninus Pius x and the early years of the third century.2 But Mr Richmon bees dha n abl shoo et wperiodo thattw e placn s,i th revealef eo d by the previous excavations at High Rochetser, there were as many as five lase whicf th ,o t h mus assignede b t e evidencth n o , potteryf eo o t , the early years of the fourth century. That seemed to justify the pre- sumption that further digging at Birrens would reveal more than the two periods distinguished in 1895, especially when it was remembered that in 1895 it was not yet a commonplace that the Roman occupation of the north of Britain was subdivided into so many distinct periods. And the fact that excavations or chance finds had shown the other outliers—High Rochester, Risingham, Bewcastle Netherby—td an , o have continue Roman di n hands t merelno , y throughou e thirth t d century, t welbu l inte fourth,oth 3 seeme justifo dt e suspicioth y n that Birrens t beeno nd lefha t unoccupied afte destructioe th r A.Df no . 196. When I turned to consider the evidence from Birrens itself, it seemed to show that my suspicion was justified. It had been demonstrated, indeed, that the gold coin of Constantius Chlorus recorded from there has no necessary connection with the presence of the Romans at Birrens, amuletseen d sinca e cut-glas ha nth s t a ei lond e ; an 4gus s beakerr fo , whic hlata t efirsa datd t eseemeha d certain, might well have reached a considerable sit t th a e y earlier time.t amonBu e 5inscription th g s there were three which, though they are not dated, should belong to e thirth d century rather tha e secondnth , namely dedicationo tw , o t s Mercury by a college of his worshippers 6 and the altar set up by the architect Amandus in honour of Brigantia.7 The attribution of the latte rthire altath do rt centur bees yha n confirme. NS . r Miller'M y db s convincing identification of the dedicator with the Valerius Amandus attested on a German inscription of A.D. 208,8 and the style of the two altar Mercuro st y clearly best fits suc hdatea complicatee th : d ligatures e abbreviatio th f the o d e moan non rara f eno nomen s firsit to t thre e letters on the other cannot lightly be ignored. Furthermore, an examina- 1 OIL, vii. 1041. 2 OIL, rii. 1043, etc. . BirleyCf 3 , Three Notes Romann o Cumberland, CWn i 2, xxxi . 137-147pp . ; Richmond, AA4,

xiii. cit. PSAS, lii. p. 219. Ibidem, and cf. p. 335 below. 5

6 OIL , vii. 1069 and 1070. l * OIL, vii. 1062. » JUS, xxvii. pp. 208-209. 280 PROCEEDING E SOCIETY , 19389 TH F Y .O S MA , tion of the 1895 plan suggested that it retained evidence of more than two structural periods, and that the inscription of A.D. 158 could not be associated wit lase themf hth o t . centrae th n I l bloc buildingsf ko , Sites IX-XV plae th , n only presents structurao tw l periods—"primary" show blacn i nd "secondary an k " bluen i t whe bu e ;"secondary nth " wallin s eliminatedgi fign , i 2 . s a , t wil i e seeb l n that ther stils ei l more thaperioe non d representedr fo , buildinge th plainlsV numbereXI yd interferan dX e wit symmetre hth y e blockoth f . Sit makeeX extremeln a s y awkward junction wit, hIX leaved an rooo sn m eve r pedestrianfo n e westraffith e tn o th csid f eo Granary XI; Site XIV is an obvious addition to XIII, and similarly blocks access to the street which must originally have run between XIII and the Granary XV; and the 1895 plan shows "secondary" walling overlying eliminatpare XIVf w to f I . e these intrusive structures, we obtain the intelligible arrangement of five buildings shown on the accompanying fig. 3: in the centre is the principia, Site XII; to the prcetorium,e th e ar t i eas f Sito t e granaryXIIIa wese d e th an , th t o t ; qucestorium, Site IX, and another granary. And an accidental discovery from elsewhere allow e suggestioth s n that this arrangement shoule b d assigned to the re-occupation of Birrens attested by the inscription of A.D. 158. A military diploma discovered at Eining in Bavaria shows 1 that in A.D. 147 a vexillation of the second cohort of Tungrians was serving in the province of Raetia, and a comparison with the Raetian diplomas suggest7 15 d san tha 3 detachmene fo15 th tr t only returne Britaio dt n between thos dateso etw t seem I . s clear that r somfo , e reason, half o f the regiment had been transferred for a time to strengthen the garrison of Raetia, where it counted as an independent cohors quingenaria; and dut t n becamdischargr i o n fo ys i whil e thern wa e du t e i me e e e somth f eo fro e completioe armyth mth n o , f theino r twenty-five years wite hth colours, and so it came to be included in the diploma which set forth the grant of the privileges customarily accorded to the soldiers qualified for honourable discharge named an ,e regimentdth e provincth n i s e which qualifiedo s consequence n On me . d ha e was, tha s ranke gapit th t n si s had to be filled by the enrolment of fresh recruits enlisted, as was usual in this period, in the province where the unit "was serving, namely Raetia. That explain occurrencee sth altan a n r o ,foun t Birrensa d f o , c(ives) Raeti milit(antes) coh(orten i secunda) TungrorumJ1 Recalled from Raetia when it was necessary for the cohort to occupy the milliary fort

GIL, xvi. 94. Birleycf ; NoteA , then o second cohort of Tungrians, CWn i seq.t , e xxxv 6 5 . p . 21 OIL, vii. 1068. 2 BIRREN S- PRIMAR Y STRUCTURES

XXIX ""H XXII ' " -J I

I__K2L XXVII J I——xix r rrxxyn M \

• •

| I I , XVt I X.XIII I

. AFTEEB . . RBARBOUJ R Pig. 2. 282 PROCEEDING E SOCIETY , 19389 TH Y F O .S MA , of Birrens at full strength, the detachment brought with it these new recruits t seemi t havo sBu t . e brought wit t hsomethini g els wells ea . Ther remarkabla s ei e feature abou centrae th t l bloc buildingf ko t sa Birrens: not only the granaries, but also the principia and the prcetorium e providear d with frequent external buttresses (cf. f slightefigo ) .3 r constructio mord nan e widely spaced e granaries thath n ni whicn i , t hi is usua o fint l d buttresses providede cas th s quiti et eBu . different with regimental headquarters and commandants' houses; we know of no other fort in Britain where such buildings are treated in this way, but in Raetia there are several instances. It seems difficult to escape

e conclusioth n thae exceptionath t l treatmen f thes o buildingo t tw e s 1 at Birrens was a by-product of the half-cohort's period of service in Raetia, e thre th e firs f eth o t n i p u thad t e inscriptioan se th t s wa f A.r>8 no 15 . trrrni

Fig . Centra3 . l block, original state. periods which the plan allows us to distinguish in the central block of buildings. It has been argued, indeed, that the circumstances under which that inscription was found show that it was still in position when Birrens •was deserted for the last time, and that therefore it 'must have been set up when the principia was rebuilt, not when it was originally constructed.2 But that argument will not survive a critical examination. The dis- cover s recordeyi e followinth n di g terms : "Som f theseo e fragments

were found near, others 3 wele inth , l thadiscoveres wa t d withi aree nth a of the" principia. In other words, the pieces "were lying in the courtyard e inscriptio e buildingth o th f i t d beebu nha ; n stil position i l n when s abandonede forth wa t s fragmentit , s should have been found fallen upo e principia, e soute streeth th nth s origina f it ho o t t r fo l position must have frone , bee th thaf o tn n i t building, wher t ecouli seee db y nb people approaching fro portae mth prcetoria. t caI n onl e concludeyb d thae inscriptioth t d beenha n re-used,s ofte e casea flawa th n s g s a ,a or flags in the paved courtyard. Analogies are not far to seek. It will 1 ORL, 61a (Sulz), 64 (Schierenhof), 660 (Xlrspring), 68 (Ruffenhofen), 70 (Gnotzheim). Cf. also 2a (Niederberg (Murrhardt4 4 d an ) Germanin )i a Superior. 2 BWS . 4782p , . 3 PSAS, xxx . 129.p . EXCAVATIONS AT BIRRENS, 1936-1937. 283

be sufficien e fragment th refeo t e tinscriptio o t th r f o s n recordine gth erection of the principia at Rough Castle, found "in a hole among other debris" in the courtyard of the headquarters building there,1 and to the dedication to Antoninus Pius from the well in the principia at Bar Hill; 2 interpretatioy an fon o r e threth f eno Antonine periodnorthere th n o s n limes neither of these inscriptions can have been set up at the opening thire olasd th f dan t occupation: indeed, though neither ston dateds ei , doubo n e therb t than erectioe etime ca th t th f bot eo t a n hp weru t ese governorshie forto th n tw si e oth f Lolliuf po s Urbicus. To sum up: it seemed clear that there were at least three structural periods represented in the remains planned in 1895, and that the first, not the last of them, must be associated with the inscription of A.D. 158; and further, that the fort was not finally abandoned at the close of the second century, but was re-occupied in the third. It happened that the Dumfriesshire and Galloway Archaeological Society paid a visit, in the summer of 1935, to my excavations at Chesterholm occasiod ha I discuso d nt an , probleme sth Birrenf so s with t . oncReids a president C it e e. h ;R suggeste r M , d that that Society should rais a fune d large enoug o allot ha tria f wo l excavationd an , invited me to superintend the work. Permission to dig was readily James Mr sd Mackiean grante r M proprietorse th y , d b e excava th d an -, tions took place in August 1936; their object was restricted to a fresh examinatio e stratificationth f no discoveo t , r wha e totath t l numbef o r structural period d beeha sn and y applyin,b g increased knowledgf o e pottery evidence, to see whether I was justified in inferring for Birrens histora y simila othee thao th rt f ro t outlier Hadrian'f so s Wall. e resultsTh , describe e followinth n di g sectio f thino s report, were

striking. As many as five structural 3 periods were found, and although there was very little stratified pottery there was enough to suggest a provisional dating of those periods: (1) to the time of Agricola; (2) and beginnin e secone th th (3o e t o )t ) th d third f e (5 centuryg o th d o an t ; ) (4 ; t seemei fourth t Bu d. essentia o securt l largeea r serie f pottero s n yi orde provido t r esurea r correlatio five th e f periodno s wit e historhth y of the Roman occupation, and there were several problems raised by an incidental examinatio e ramparth f gatewayd no an te for th t f whico s h called for fuller examination. I was fortunate enough to persuade Mr I. A. Richmond to join me in the direction of this further excavation, which took place in July 1937, with the aid of generous grants from the

1 BWS . 2282p , . 2 RWS . 2792p , . preliminarA 3 y report appeare Dumfriesshiren di Gallowayd an Transactions, . 157—170pp . xx ; reprint . 1-14pp , . 284 PROCEEDINGS OF THE SOCIETY, MAY 9, 1938.

Council of this Society and from the Trustees of the Haverfield Bequest, as .well as fresh contributions from members of the Dumfriesshire and Galloway Society. e eventth n fula I , l enough rang potterf eo obtaines ywa givo dt ea definite answe problee th o rt m tha taked Birrenso t ha tgreaa s nu d tan , dea structura e ligh f shes lth o n tdwa o l history ramparte th f o additionn i ; , fresh problem f whico s l aroseal h t coulno , solvee db d befor e seconeth d season's work cam a closet i wil d som necessare o t ey b lan , da e r fo y further work to be undertaken, to complete the investigation of the points which still remain obscure. But as we are now in a position to answe e originath r l question t seemi , appropriatn a s ebefory timla o et this Societ yworrepore a th kn o tdon 193n ei 1937d 6an .

II. THE EXCAVATIONS OF 1936. The first season's work comprised examination of the stratification on two sites, the west end of XIX in the retentura and part of VIII in the prcatentura of the fort; the cutting of a section through the west rampart, opposite Site XIX; and a partial re-examination of the north and west gates. The last-named operations were disappointingly inconclusive. firse th t n instancI t trenche cu gatewaye e th e w n si s merel alloo yt s wu o pict p pointu k s from whic e coulw h d measur e positioth f of f e o n Site XIX d selecte, whicha r e particulahw fo d r attention because th e 1895 plan showed that a minimum of digging had been done there in that year. But it soon became clear that the gateway structures planned in 1895 had been wholly removed after the planning of them had been completed. At the north gate we were unable to find any masonry in position fullea d ran , examinatio 193n n i s hardl 7wa y more successful; but at the west gate we did succeed in finding a gateway, though it was not the one found and planned in 1895. The West Gateway.——The accompanying diagram (fig. 4) gives an enlargement from the 1895 plan, together with a plan, to the same scale, of the gateway that we found in 1936. The difference is not merely one of dimensions f materialso t e bu 189, Th .5 gatewa s describei y s a d being of good masonry, "the stones, of various sizes but generally small, being squared weld an ,l fitte bonden i d d courses" ;contrasn i 1 thiso t t , .the masonry foun n 193i d e s extremrougth 6wa n d i seemeh an e a d complete puzzle, until Mr Richmond explained it convincingly as the rubble filling of a timber framework. Originally the gate-passage must have been abou fee8 1 t t long, with four upright beams recessed into each 1 PSAS, xxx. p. 103. EXCAVATION BIRRENST A S , 1936-1937. 285

WEST GATE 1895

WEST GATE 1936

r~i POST

20 =d FT

Fig. 4. 286 PROCEEDINGS OF THE SOCIETY, MAY 9, 1938. side at the ends and at one-third and two-third intervals; the outer thir bees dha n wholly remove stone-robbery db excavatorsr so . Extend- ing southwards from the inner end of this gate-passage was a narrow wall, which returned eastward feet4 1 t ; a s withi e anglnth formeo es d wa conceivabls i heartha s t i d an , e tha have w t e her e remaineth a f o s guardroom attached to the inner side of the rampart; but we did not complet examinatioe eth thif no s area, which will hav receivo et e atten- tion anotheo n r occasion t I "wile remembere. b l d tha a somewhat t similar structure was found in the same position relatively to the "secondary" north gate'in 1895,-1 but we did not succeed in establishing the relationship between the 1936 west gateway and any of the levels found elsewhere in the fort, and the date of this timber and rubble gateway remains a matter for conjecture. We were careful to leave the northern side untouched tha o e connections , th t s between passage-wall, turf rampart, and internal levels can still be established. It will be convenient to describe the section cut through the west rampart, farther north, opposite Site XIX West, after recording the results obtaine than do t Sitsitd ean VIII.2 Site XIX West.—Fig. 5 gives a record-plan of the structures found' on this site, together with separate plan stone eacr th sfo f eho buildings; notes follo five eacn wth eo f hstructurao l periods. Level I.—Four post-holes, insufficient to indicate any plan, and a short lengt f sleeper-trencho h , compris e e evidencwholth th e f o e n o e this site for an original timber building; there was only one piece of pottery at this level—part of the flange of a Samian bowl of Curie's typbelow)1 1 (fig, 1 e 26 . , whic e timh th f s latmigh o a es a ee b t Hadrian. Level II.—The first stone building had a doorway (whose width we did not ascertain) in its south wall, 15 feet from the west end; there was a small patch of flagging inside that wall and in the doorway itself (cf. fig. 8), but otherwise the floor was of clay, spread over and securely sealing the post-holes of Level I. The only pottery from this level is illustrated below (fig. 26, 2 and 3); it appears to belong to the period A.r>. 120-160. Level III.—As reconstructed d rathee buildinth ha , w r thinnegno r walls addition I origina.e th o nt l doorwa soute th n hyi wall, whics hwa retaine usen di doorwaa , feey6 inche6 t s wide,3 wit hflaggea d threshold which showed sign f considerablo s e wear s inserte e weswa , th tn d i wal l (cf. fig. 9); a rathether s ewa r larger are f flaggingao larga d ean ,fla g

1 PSAS, xxx. pi. Hi., facin . 102gp . 2 Below . 293 p ,. als . 302-306cf ;opp . 3 Cf. the 10-foot doorway in the stables at Halton, AA4, xiv. p. 164. EXCAVATION BIRRENST A S , 1936-1937. 287 servin threshola s ga e sout th n hdi doorway, abou inche6 t s above eth floor of Level II (cf. fig. 7). No pottery was found at this level.

V P- HV I l Il

j_____|____I 10 15 2O FEET

BIRRENS - SITE XIX W.

RECORD-PLAN & PERIOD-KEYS

E.B.

Fig. 5. Level IV.—In this buildinperiow ne erecteds da e sam gwa th en o , general lines as its predecessor but not on the same foundations; the 288 PROCEEDINGS OF THE SOCIETY, MAY 9, 1938. south wall of the new building was almost wholly south of that of Level III, whic mainls hfloowa w yne r coveree (fig flage th , th f .6 o sy db

Pig. 6. Site XIX West: flagging and south wall of Level IV.

Fig West. SitX 7 . eXI : flaggin soutd gan h wal Levef o l l III. and cf. figs. 7 and 8). There was no longer a doorway in the south wall, and only a narrow one at the west; nearly half the floor-space was now EXCAVATIONS AT BIRRENS, 1936-1937. 289

flagged (fig . occupation-deposio .n Ther6) s ewa f potteryo t t frobu , m

Pig. 8. Site XIX West: flagging and south wall of Level II.

Fig. 9. Site XIX West: inner face of west end, showing remains of Level II wall covered by Leve I threshol courseII lo tw Levef d so d an walling V I l . among the debris overlying the floor came the outbent cooking-pot rim (fig. 26, 4), which can hardly be earlier than the close of the third century. 290 PROCEEDINGS OF THE SOCIETY, MAY 9, 1938.

Level V.—The 1895 plan shows the south-west corner of a "secondary" building overlying the '' primary'' walls of this site; the latter are clearly those of our Level IV; but no traces of this later structure were found survivin were w fin o 1936n egt s i dA elsewher. t Birrensea remaine th , s had been badly robbed after their planning had been completed. In fig. 5 the walls of this level have been plotted in by enlargement from earliee th r plan. The scarcit f potteryo y inside thi parn i s s sitt compensateewa r dfo by the discovery of two vessels, securely stratified, in the alley-way to themf o e juga bee,d On ha , n norte . depositeit th f ho allee th t ya n di Level III, most of it being covered by the north Avail of Level IV, which had to be removed before all the pieces could be secured; the other, a large mortarium, came from the same level, but further to the east, where the wall of Level IV had been wholly destroyed. Both pieces are illustrated in fig. 27 below; they appear to belong to the close of the second century thud ,an s give usefu,a l terminus post beginnin e quernth r fo g of the period represented by Level IV. t Suci was s a h , then e stratifieth , d pottery from WesX SitXI e t allowe e followindth g provisional datine successivth f go e periods:— Level , timberI building : Hadria t latestna l (Samian bowl, Curie, 11). Level II, first stone building: A.D. 120—160 (platters best paralleled on Hadrian's Wal than li t period). Level III, second stone building: A.D. 160-200 (jug and mortarium, broken at the end of the period, assignable to circa 200). Level , thirdIV building : A.D . 200—300 (cooking-pot n overlyini , g debris, assignabl circao et 300). Level fourth, V stone building : A.D onwards0 .30 . But it was clearly desirable to obtain a more extensive series of pottery, and accordingly attention was transferred to Site VIII in the prcetentura, where the surface indications promised reasonably intact stratification and analogies suggested that we should find buildings that had been occupie s barracksda , rather tha e stablenth s which repree seeb mo t - structuresentee th y db Sitn so e XIX. Site VIII.— Heree 189th 5 plan make distinctioo n s f periodno n i s complea wallsf xo t firswhica n i satisfactor,y s i t an t difficul hi e se o t ty indication of a coherent plan. But as soon as the possibility of more thaperioe non d being represente s entertaineddi e probleth , m becomes 1 In the preliminary report it was assumed that this level belonged to the period of Agricola, .because of the presence of a pre-Hadrianic rim in the series of material from- Birrens preserved in e Dumfrieth s Museume discoverie vien th i f t wo bu ; s mad 193n ei t 7seemi s wise suspeno t r d judgmen thin o t s poin below)5 t34 (cf.. p . EXCAVATIONS AT BIRRENS, 1936-1937. 291

less acute n enlargemena . s i Fig0 1 - f thao t t o whicplant e n o th h, excavatio 193f no 6alss i o plotted studplae A th . n f y o suggest tha) (1 s t

CD Ol

O Z CD O a

ro ~U O)

^ oj ro

UJ o_ t en 1o

S ° en i Z O I UJ ? O f£ o UJ 1- o o ce uj 55 UJ UJ .MIi o G 5 I/) 1_

at the east end we have a series of independent huts, separated from each other by narrow alley-ways, as in the Constantian barracks at 0 2 VOL. LXXII. 292 PROCEEDINGS OF THE. SOCIETY, MAY 9, 1938.

Birdoswald and High Rochester; and (2) that in the first instance the occupies sit lono narrod ewa tw gan y dwb barrack buildings, separated a stree y b t (cf. fig ) whos.11 e outlin stily discernee b ema l spitn di f eo extensive rebuilding, whic involves hha disappearance dth streee th f eto and the partial obliteration of the original plan.1 smale Th l portio site th e f reopeneno 193n di choses 6wa n morr efo s promisit stratificatiof eo npossibilite thath r nfo y tha t mighi t t throw light on the structural changes suggested by a study of the published plan; but it proved to contain evidence in support of our second infer- ence, for a fragment of wall, not shown on that plan, was found in westward continuation of what we'have taken to be the northern wall of the original southern barrack (fig. 10), and that wall proved to be structurao earlietw y b r l period aboud parallee an s t on ti tha e o t lnth

STREET

Fig. 11. Conjectural first lay-out of Site VIII.

6 feet farthe r e latter'northth r fo s; foundations reste a cobble n do d roadway, which extended over the remains of the former. This was not the only evidence for a series of structural periods comparable to what had been found on Site XIX West. Within the .area examined the following sequence was observed:— Level I.—A series of post-holes, cut into the subsoil, seems to belong to wooden buildings alined diagonally to the existing fort; no pottery was found at this level. Level II.—The cla ye souther th floo f o r n barrac retaio k(t na con - venient term) overlay the post-holes; it was covered by a deposit whose maximum dept s abouhwa inches3 t , containin ga quantit f burnyo t matter, including many pieces of partly charred wood. There were uninformativw onlfe ya e scrap f potteryo s t therbu , e were abou0 20 t fragments, som f theo e m distorte y intensb d e o heattw r ,o froe mon glass vessels, together wit e interestinth h g grou f bronzeo p s thae ar t described below, p. 337. 1 It is possible, on the other hand, that the arrangement may have been like that at Birdoswald in the second century, with a narrow store-building lying along the via prcetoria and a standard barrack behind it; the 1895 plan does .not preclude the possibility of such an original arrangement. EXCAVATIONS AT BIRRENS3 , 29 1936-1937.

Level III.—The burnt deposit was sealed by a fresh clay floor, merg- ing int cobblee oth d roadway which covere norte dth hbarracke walth f lo . This floor was only a few inches below the turf; the associated structural remains had been badly disturbed—for example, the wall running north and south across this area, shown on the 1895 plan, had been wholly t possiblno s removed telo wa et t li whad an , t for bloce mth buildingf ko s had now assumed, or which of the original walls were still retained in use. Nor was it possible to distinguish for certain between the pottery e disturbeth flood w an ne depositer d e materiath n o d l fro e overmth - lying Level IV (see below) that had been removed; the pieces selected fo r therefore illustratioar 3 3 d e an describefigs n i 8 2 . comins da g from Level III +. But while some of this pottery must be assigned to the third century 7, (fig mos, ;t seem i 28 1) fig.f , o t belon o .s33 lattet e th o grt part seconde oth f . It will be seen, then, that Site VIII retains structural evidence of a timber building, Level I, followed by a first stone building, Level II, whose occupation ended in destruction by fire; the second stone period, Level III produces ,ha d potter lattee th f yrsecono e parth f o td century; an thirda d stone period, attestes Levei , IV lthird-century db y pottery, and by the more northerly wall, already referred to as being two periods later than the original southern barrack. If the interpretation put forward e structureabovth r fo e sf thi o planne se easd th siten t t eda is correct, e thebelony Constantiath yma o t g n perio d represenan d t Leve , completinV l ga serie s simila o tha t rSitn o t e XIX. Thers i e little doubt that a careful re-examination of the whole of Site VIII would allo e productiowth a tolerabl f no y complete seriee th f planso st bu ; time and means available in 1936 did not allow further excavation, which must be deferred until another occasion. The Section through Weste th Rampart.—Fig. give2 1 sdiagrammatia c representation of the features observed in this section which, as it will be seen presently, differed in material particulars from those cut in the east and north ramparts in 1937. Three periods were noted in the body ramparte oth f firse Th t ."wa s represented solel stria y f turfworkypb o , 11 feet from east to west, laid immediately on the subsoil, and under- lyin e foundatiogth e seconth f o nd periodf cobbles o e latte s Th wa .r , wit hlarga e outer kerb measured an , fee3 d1 t from eas westo t t ; from the kerb to the ditch was a sloping berm, 7 feet wide; the ditch itself slope angln a abouf t dea o degree0 horizontale t4 th coule o st t w t dno bu , get as far as its centre, as there is a small sike running by the side of the field-hedg t thiea forse sidth t f trenchwhicr e o limi a ou t o ht se . About fooa t bacse t k fro fooa kerbe md th t an abov, , cameit foundatioea f no BIRRENS - 1936 SECTIO F WESO N T RAMPART

LEGEND

TURF & EARTH E3

TURF ^

GRAVEi H L

COBBLES M SUBSOIm a L SCAL F FEEO E T E.B.

Pig. 12. EXCAVATIONS AT BIRRENS, 1936-1937. 295

flags fee3 , t only from eas west,o t whicn 1o h reste dwell-definea d cheek of laid turf, which gives the third structural period; as Mr Richmond points out below, this seems to correspond to the re-facing of the north rampart sectiono , 1937 n i westernotee tw t th e cu sth n .d i f no Belo ture wth f cheek withirampartd e th an , , formens it wa mixef do d earthd turan f ;inneo n ther s r foundationwa e a widt thers bu e, wa e inner cheek of laid turf, whose lower part was not observed sufficiently carefull relationshis it r yfo subsoie recordede th b o pt o lt thao s , t musi t t remain uncertain whether or not it represented a rearward addition to rampare th Periof to . dII Withi n this inner cheek came three successive levels of the intervallum road, each extending farther west than the one belo , surfacewit d "witregrettee b ho t gravel s i d t I tha relationshi.e tth p between these three levels and the successive buildings on the adjacent establishede b coulX doub Sito t n XI eno de t b ther than bu ; ca et further work near by, in the light of the experience gained in 1937, would enable a relationship to be defined.

EXCAVATIONE IIITH . 1937F O S . Apart from the further examination of the rampart structure, dealt Richmonr witM y hb Section di belowV nI somd an , e trenching neae rth granary teso e deptt , , th t Sit f stratificatioho XV e n survivin than gi t forte parth f , o twor 193n k i concentrate s 7wa parto Sittw f so n edo XXII, where surface appearances seemed to promise that we should find a considerable depth of stratification. In the event, we found that we should have been better advised to select a building in the eastern part subsoie th forte r oth f fo ,l prove risine b o gdt more sharply towarde sth north than the present surface suggests, whereas it falls towards the east n consequencei ; e maximuth , m undisturbed deposit e immedili s - ately inside the east rampart. Site XXII East (fig. 13).—Here conditions were particularly dis- appointing e sitth e r provefo , havo dt e been drastically denudedt No . only had the upper levels been almost wholly removed, but the north wall and all but the southernmost 12 feet of the east wall of the earliest stone building had gone. There was a partition wall running, northwards from the south wall, 25 feet from the east end, and immediately west of it there was a 3-foot doorway, which had been blocked up in a second period (cf. fig. 14), when the partition wall had been demolished and covered by a rough tumble of stones. Within the room to the east of the partition the only surviving stratification was in the centre, where 1 Cf- the outer kerb in the section through the east rampart, p. 302 below and fig. 18. RAMPART OF ANTONINE FORT_: _I_8_FT. THICK_ _ .... .,, 7_ 3 7\ z

Fig. 13. EXCAVATIONS AT BIRRENS, 1936-1937. 297

a secondary hearth overlay a deposit that contained the greater part of a deep flat-rimmed platter (fig. 31, 13), part of which may be seen in the photograph reproduced as fig. 15. In the second period on this site the building had been extended southwards across the street on to which the doorway already mentioned had opened; the "primary" wall, e show189th n 5no plae uprighth n s (figa ) t 2 . strokinverten a f eo dT

Fig. 14. Site XXII Bast from the south, showing the walled-up doorway and demolished partition wall. lying across this street doubo n , t belong e samth eo e t speriod th d an ; black occupation-layer that marked the floor of the extension yielded the beaker rim (fig. 31, 9) which may be paralleled in third-century deposits on Hadrian's Wall. Careful search showed that there wer post-holeo en s underlyine gth barrack or the street, and a discovery to be recorded presently explains why there e timwerth er e beinfo none t t gwili bu enoug;e y b l sa o ht that this site, with onlsurvivino ytw g stone periods nond timberf an , eo , presente completo s d a contrase e resultth o t st from SiteWesX XI s t and VIII thaprobleme th t i onl t o t y d sserve witad o hdt whice hw were faced. Accordingly we turned to another area, Site XXII West, 298 PROCEEDING E SOCIETY , 19389 TH F Y .O S MA ,

nearl fee0 y10 t farther wesd somewhaan t t farther south, where there "were surface indications of rather deeper stratification. Site XXII West.—Here we were rewarded with an instructive if complicated serie f levelso s , whic t hwili conveniene b l o describt t n ei the order of discovery rather than in the historical order observed hitherto.

. FigSit15 e. XXII East; secondary hearth portiod an , flat-rimmea f no d platter.

(1) Immediately belo e turcame wth w f e upo nwala l running east and west, with a southward partition 8 feet long with squared end, mark- ing the north side of a doorway; there was nothing left of its south side and the southward continuation of the partition; to the north, at the west side of our excavation, there was another partition-wall, interrupted by a doorway rather under 3 feet wide. These walls are shown in solid foune W blactraco dn fig. n keo 13 . survivin e floo th thif o f rg o s build- possibls i t i t ingebu ,tha t pottere somth f eo y topsoie founth y n di ma l have come from it; and there was a rough buttress against the south e east-wessidth f eo t e bodwallth f whicrimso yn o i , foune tw h,w e dth . 12 d an 1 1 , fig31 . EXCAVATION BIRRENST A S , 1936-1937. 299

(2) Interrupte e wallth sy db jus t referre , whic to dd bee t hha cu n into it, there was a layer several inches thick of burnt matter, among which quantitthera s ewa potterf yo y (fig , 1-8.31 ; fig. 32, 2-4). Within the area examined there were no structural remains that could be associated with this layer, which overlay a spread of debris that is clearly shown in fig. 16.

Fig. 16. Site XXII West fro weste mth , showin late gth e east-west wall founde layea n debrisf do ro , and the clay and cobble footing and rough stone floor of the earlier northern barrack. (3) The latter spread in turn covered the remains, reduced for the most par o theit t r clad cobblan y e footingso stontw ef o buildings, , running east and west and separated by a narrow alley-way; there were two partition-walls in the southern building and one in the northern seee b fig n ni e roomo (t .th 16) formed o s an ,floor d dha s eithe claf o r y rougf oo r h flagging. pottero Thern s eywa associated with this level. (4) Lowest of all, reaching a maximum depth of about 6 inches belo e latwth e east-west wall , a deposither s f eturfworwa o t k exactly comparabl lowese thath o et f o tt perioe sectioth t througn di ncu e hth west rampart in 1936 (p. 293 above), namely, laid directly upon the sub- soil, without any foundation. This "was clearly the remains of a rampart; 300 PROCEEDING E SOCIETY , 19389 TH Y F O S. MA ,

searcd norte t quickli an th f ho ho t y reveale associatee dth d ditch, •which d beeha n fille when i de buildingnth Levef so wer) (3 l e erectedt a d an ;

t acros Cu earle Pig. sth y17 . ditch, easter nretentura,e sidth f eo fro north-west.e mth

this point there •was als east-wesn oa t wallatea f o lr period overlyin, git betwee ne ditch-fillin whicth d han g ther seales ewa e smal dth l grouf po pottery illustrated as fig. 30 below. Further cuts were made to establish the line of the ditch, the first at 10 feet farther east, the next opposite the west end of Site XXII East, and a third (fig. 17) in the eastern half of the retentura; these showed that it had run approximately parallel existine th o t g north rampar nearld an tfee 0 y5 t sout, measurinit f ho g EXCAVATIONS AT BIRRENS, 1936-1937. 301

fro originas mit l inner edg ditco t possiblet no hs centre obtaio ewa t t I n. as full details as we could have wished, for the discovery was only made in the last week of the season, and we were hampered by water whenever dowg wedu n into the"ditch (c/. fig. 17). This level produced, belo e centrawth l southere rooth f mo n building of Level (3) e greate,th ra large sid f th par eo ef o tSamia n plattef o r form 18/31 fracturee edgee s surfacit th th ; f d so beed ean badl ha so n s y damaged by the action of the soil (which at Birrens generally has this effect on Samian ware) that a drawing seemed out of the question; but thicknese comparativele th fabrie th th d f can so y heavy profile preclude possibilite th data f yo e earlier tha time n th Hadriaf eo n being suggested for the vessel, and it might well belong to the Antonine period. The significance of this discovery is plain for all, that fuller informa- tion mus e soughb t y furtheb t r digging. Comparison wit e weshth t sectio 193n i t 6ncu show s tha have tw e her norte eth h rampart correspond- e lowesthree th th in f o egt o t that were noted Richmonr thereM s a ; d points out below, the sections cut through the visible north rampart, and through the east rampart at a point farther north than the early ditch, produced nothing comparable; in other words, the earliest west rampart returned eastwards 50 feet or more south of the eastward retur e lattevisiblew th nno d r an represent, a northwars d extensiof no forte th . Evidence for the date of the extension is provided by two deposits of pottery; the first is the Samian platter already referred to, which shows thae occupatioth t e unenlargeth f no t n coma dno o et ford di t beford en e tim th ef Hadrianhavy eo ma e d lastean , d int e timoth f o e Antoninus e Piussecon a Th grou .s i df potter po y found underlying e visiblth e north rampar d 29)an ,3 t whic 2 (figs, 22 h. includes pieces ranging in date from circa A.D. 120 till about the middle of the second century t doeI t see. no sm unreasonabl suggeso et t thaextensioe th t n was the work of Julius Verus, in the course of the reconstruction already attested by the inscription of A.D. 158. In that case, the lack of corre- spondence between the levels observed on Site XXII and those on Sites XIX and VIII becomes less of a problem; for the pottery already found entitles us to postulate at least one period, and possibly two, in the occupation of Birrens before that date. But this question, for the satis- factory solution of which the evidence is still too scanty, must be left for discussion in the concluding section of the present report; at the moment t wili sufficiene b l noto t t e tha absence th t post-holef eo Sitn so e XXII is adequately explaine e facth t y tha b de building th te enlargeth n i s d fort were all of stone. 302 PROCEEDING E SOCIETY , 19389 TH Y F O S. MA ,

remaindee Th wore th f kro don 193n ei 7describes i Richmonr M y db d followine th . n V i sections o d gtw an V I ,

. RICHMONDA RAMPARTE . I TH . y B IV .. In 1895, when the first sections were cut through the ramparts at Birrens founs wa dt i , difficul interpreo t t t eve maie nth n featuree th f o s structure. Doubt was expressed as to whether the stone bottoming noted below the rampart was intended to hold masonry or earthwork. To-day possibllittls a i t t i ,e ge furthero et , largelbasie th observa f sn o yo - tion in the light of comparative material. Many points, however, still remain obscure, and it is just as evident as in 1895 that much further •work mus carriee b t t beford ou rampar e eth t structur completele b n eca y understood. The East Rampart (fig. 18).—This sectio ne simplest th prove e b o dt . originae Th l rampart fee0 2 , t wide composes wa , cora f mixef do eo d turf and upcast, retained between two very massive turf cheeks. The toe of the front cheek was set upon a 3-foot kerb of rough stonework, topped with e rearwarclayTh . d chee s supportewa k d upo a nbroa d kerb, 8 feet wide, and was bonded into the material of the core by a long tongue of turfwork. Both cheeks -were linked at the base of the rampart by a bedding of turf. While, however, weathering had destroyed the original face of the front cheek, the profile of the rearward cheek had been perfectly preserve additioe th y masda b f nmixef o so d earth, addin t leasga fee0 1 t t to the back of the rampart. The rearward extension of the rampart covered an early intervallum poine th road t t A wher. e sectioeth takes nwa n (c/. fig , wher.21 e eth positions of the sections taken in 1936 and 1937 are marked) the additional material had also been cut back to receive a well-built oven of the same type as those discovered close to the east gate in 1895. On a level with this ash-strew ovenw ne a , n layer began insid nots i fort e t I e,th .how - ever, clear what the relation of the oven to the extended rampart may be. This depends upon whether a level intermediate between the early intervallum ovee th roand existsdan . In this section, two clear stages of construction can be detected, and extension a e sectioth f no n would soon revea e relatioth l n betweee nth rampart and the buildings of the fort. It may be added that this section was cut north of the limits of the earlier fort (c/. p. 301 above). The North Rampart, West Section (fig. 19).—The same sequence is visible here as in the east rampart. Stone kerbs at back and front carry originan a l rampar fee9 1 t t thick. Here, however rearware th , d cheef ko EXCAVATIONS AT BIRRENS, 1936-1937. 303 304 PROCEEDING E SOCIETY , 19389 TH F Y O S. MA , EXCAVATION BIRRENST A S 5 , 30 1936-1937.

turfwork is missing, and the kerbs are considerably thicker. They are laid in occupation-earth, covering a bedding of clean sand, spread upon marshy ground. The pottery from this deposit is considered below (p. 310 gived 29)d , an Antoninn an figs, sa 2 2 . eramparte datth e o et Th . front cheek of turfwork is pierced by two large elements of timber, the foremost an upright post resting upon the kerbing and held in position by turfworke th hindmose th , horizontaa t l beam running parallel wite hth front of the rampart. Behin originae dth l rampar extensioe th t presents ni massiva d an , e featur s beeha en delved away fro s backmit . This missing featurs i e probabl e stonyth e revetment which occur t thia s s e easpointh t n i t section of the north rampart. It should be remarked that the turf cheek holding the timber work extend r beyonfa o kerbine s dth suggeso t s a g t very strongly that this cheek is not original, but is a refurbishing of the front of the rampart. Secure evidenc r sucefo hchanga obtaines ewa sectioa 1936n d i n i ,f no wese th t rampart, wher additionae eth l chee buils kwa t upo levellee nth d remain earliee th f so r front above)fig2 d 1 ker,. an w wit5 bne .h29 a (c/ . p . The North Rampart, East Section (fig. 19).—A third section was cut 14 feet west of the north gate. This revealed, as farther west, the two stages of construction. The early rampart, however, contained three horizontal beams, paralle fron e closd th an to e t l together, about midway throug ramparte hth . Thes presumable ear y connected wit htowea t a r the gate, of which the foundations were carried upon wooden cradling. Such a tower might be expected to go with a gate resembling the west gate discovered in 1936, a half-timbered structure with rubble infilling remarkee b (c/y . ma fig above) d4 t i . tha d lattee th tan ; r gate e fitth s early 19-foot rampart. The best parallel is the rampart of Saalburg II.

The rearward extension of the rampart here retained one course of its1 back revetment, and exhibited also a horizontal beam, passing through e rampar degreesth 5 n angl4 a f t d evidentlo ea t an , y representinga diagonal frone stayTh t. chee turf ko f extende t leasda fee6 t t beyond frone th pierces uprighttn a wa kerb y d db an muc,n i , same hth e position as in the west section. A third period in the construction of the rampart is represented by a revetmen t righta t anglerampare th o st t (c/. fig. 20), retaining t wouli s a , d rama f o platforr po d seemen stairsr e buildere mth fo , Th . thif so s work, at a considerably higher level, have removed all but one course of the back

revetment of the extension, in order to bond in their earthwork. 1 Jacobi, Saalburg JahrbucTi, iv. pp. 7-12. 306 PROCEEDINGS OF THE SOCIETY, MAY 9, 1938. Conclusion.—In conclusion notee b y d t impossibl thama ye t ^s i t tsi a e to discriminate accurately between the extension of the back of the rampare refurbishinth d e frontan tth f o .g These operatione b y ma s either distinc r contemporaryo t accurato N . e relationshi t beeye ns pha established betwee structure s nth i rampare t gatess i th it f d eo d an , an t

Fig. Lat20 .e revetmen t righa t t anglenorte th ho st rampar e easterth t a tn section foree th -n i ; ground is a length, interrupted where the side-stones have heen removed, of the water-channel.

doubtful whether the previous excavators have left enough in position mako t e suc hdefinitioa n possible. Finally denudatioe th , norte th f hno end of the site makes it impossible to work out the sequence of building- period connection si n wit norte hth h rampart. othee Thisth n r o ,hand , shoul possible b d e behind eithe e eas r th wesr o t t ramparts d offeran , s the remaining chance of associating the history of the buildings with that of their defences.

V. THE AQUEDUCT. By I. A. RICHMOND. The first trace thif so s structure were detected behin wese dth t sector norte oth f h rampart, wher ker e curioud eth ban s stone structure, noted EXCAVATION BIRRENST A S 7 , 30 1936-1937.

in 1895, were re-examined t sooI . n became plain tha so-callee tth d kerb was a continuous channel, formed either of large upright stones set on edge (and sometimes squeezed together) or of massive facing stones laid forminf i flat s a , side draina gf th e o . The stone structure box-lika prove e b o dt e tank mad verf eo y well- fitted red sandstone slabs set in clay. The bottom was flagged, but the front flag had been torn out and was tipped on end. At the sides, where e drain-likth e structure reache tanke dth e flag th , s were notchedf i s a , for inflo d outflowwan . An interpretatio e functioth f no thif no s structure would have been difficult if a close analogy had not been forthcoming. At Corbridge,1 in the same summer, a system of underground tanks closely resembling this one in type, though larger in size, had been discovered in connection wit distributioe hth watef no r fro fountaie mth aqueduct-heade th t na . While these, however, were fed from carefully built stone conduits, the Birrens example is fed from a rather clumsily built duct, with strong sides bottomo n t bu .explainee b Thin ca s d only upo assumptioe nth n that the duct at Birrens originally held a pipe-line, for such a duct as remained would effectively shield eithe leara d pipwoodea r eo n conduit. This interpretation however, demanded a source of supply entering s nort it fore t th a ht end, whenc watee eth r migh distributee b t l ovedal r the site; and levelling soon showed that the only possible line of entry was along the cause-way leading across the ditch-system to the north gate. A trenc therefors hwa t acrosecu causewae sth fee0 y4 t ramparnorthe th f o t face. This revealed, in the centre of the causeway, an underground channel composed of large stones in which a semicircular gutter had been cut, while their tops were covered with large fla t vern i slab t y se sstif f clay. Wate stils lrwa runnin channele th n gi whicf o , cross-sectioe hth n was 10 inches wide. The channel discovered exactly resemble e visiblth s e aqueduct a t Corbridge, of which a few cover-slabs are now in position, and recalls very closely Bruce's description of the water-channel supplying Brem-

enium, which entere soute th y hdb causewa forte th .f yo e Ther 2 b n eca no doubt as to its purpose. It may be presumed that the supply arrived a distribution-tan t a k behin e nort th ds thenc hwa gateed carriean , d throughout the fort. Important buildings would have their own supply. The rank and file drew their water from such dipping-tanks as the example discovered behind the north rampart. There can be little doubt that the system, or its like, was originally

1 AA4, xv. pp. 253-54. 2 The Roman Wall, edn. 2, p. 301. 1 2 VOL. LXXII. GO f >E) . \ \i I x j L \ \ & ,\\v O N ^ ,\^ 5° (K hi

I

\ N^- ;\^\\NX

P B P- n O C >\^ 8= i£> S N^^\ \\ \ ^\\ .\^\ EXCAVATIONS AT BIRRENS, 1936-1937. 309

much more common in permanent forts than the known examples would suggest. It is obvious, especially in cavalry forts, that the solitary principia,e welth n i l ofteno s source citeth s f dsupplya eo , cannot have convenience th t wholme e th f eeo garrison constanA . t supply continually flowing from tan tanko k t delivered an , d below ground, "wher presencs eit e remained unknown to the enemy, would ensure that the everyday needs were met upon a scale which conforms to Roman standards of lavishness.

FINDSE TH . . VI followine th n I g pages wil foune lb d illustration descriptiond san s a f so manobjecte th f yo s found—decorated Samian ware, other pottery, glass, bronze objects inscriptionn a , coins—ad an , claia d mpublicationo t sha , eithe s comina r g from stratified deposit r becauso s f theieo r intrinsic interest. Most of the material does come from stratified deposits, and for that reasoseemehas nit d bes describto t potteryethe , which formthe s , mainlbulit f ko y accordin deposite th o gt found s whicn si wa t h,i rather than in the manner of a museum catalogue, type by type. objecte Th s found durin excavatione gth f 193so 6 have been presented by Mr and Mrs Mackie to the Dumfries Museum; the destination of those found in 1937 has not yet been decided, though it is probable that they will be divided between the National Museum of Antiquities and the Dumfries Museum. It seems necessary, therefore, to note, in the case of each item easiee yeae b found th s whic,n y i ro t r wa tha o ma t s h i ,t i t identify individual pieces in time to come. Occasion has been taken to include notes on a few of the potsherds found during the earlier excava- tion t Birrensa s n suci ; h cases e Nationath , l Museu f Antiquitiemo s reference number takes the place of the date 1936 or 1937.

. DecoratedA Samian Ware. It has recently been suggested * that Roman regiments received their supplies of crockery from "some central authority," which presumably made large purchases direct from the manufacturers. It is hardly necessar enphasizo yt e that therevidenco n s ei supporn ei thif o t s view, when evidence to the contrary (of which Birrens can supply its quota) is so abundant. In the first place, the assumption that pottery was issued to regiments, and owned collectively by them, is belied alike by what we know of the organisation of the Roman army 2 and by the frequency with 1 By Sir George Macdonald, "The Dating-value of Samian Ware," in JRS, xxv. p. 197. For example, there is no evidence of such a practice in the Egyptian papyri in which, incidentally, private th 2e ownershi l kindal f equipmenf po so attesteds i t . 0 31 PROCEEDING E SOCIETY , 19389 TH Y F O S. MA ,

which vessel foune sar d bearin e namindividuan ga th f eo l ownero tw ; examples are illustrated below, fig. 36, 1 and 2, and several were found earliee th n i r excavation t Birrens.sa Moreover, block purchase should

result in far greater 1 uniformity in the material, on a military site than obtains in a town, supplied by the ordinary channels of competitive e attempth trade t mako bu t t; e suc hdistinctioa n between fortd an s town ss clearl i 2 y tendentious o greaten ; r uniformit e observeb n ca y d on military than on civil sites, and the present small series of decorated Samian from Birrens provide admirabln sa e exampl e varietth f eo f yo sources on which troops in garrison at a single fort could draw:3 Nos. 1 and 4-7 below come from Central Gaulish potteries whose products are regularly represented in Britain, but the East Gaulish or German bowls, Nos. 2, 3, and 9, have but few counterparts on civil or military sites in this region. On the assumption of block purchase, they had no business finto dBirrensto thei purchasway rbut ; individuaeby l soldiers from negotiatores cretarii supplie n ampla s e explanation. Ther a furthe s i e r point, however, which it seems worth while discussing, particularly in occurrence vieth f wo t Birrenea bowa f so l fro Samiae mth n potteriet sa Trier. Oelmann has observed 4 that in Germany the products of Trier hardly made their way farther south than the Main, whereas the far more extensive potterie Rheinzaberf so n were abl gaio et markenshare a th f eo t eve Trien ni r itself. Decorated bowls from uncommoo s Trie e ar r n ni northe th Britaif o permissibl e b ny thama t ti suggeso et t that they came, not in the ships or crates of pottery merchants, but in the baggage of individual soldiers or civilians. In that case, No. 9 below might well have come to Birrens with someone, like the architect Amandus, who had been serving in Germany before the British expedition of Severus; the Hadrianic , migh2 piece . tNo , equally have been n broughtma a y b transferred from Germany under Platorius Nepos, and No. 3 (to Avhich ther severae ear l parallel t Corbridgesa soldiea y b )f Lolliu o r s Urbicus. smallnesss it l al r Fo , then groue ,th decoratef po d Samian fro recene mth t excavations presents features of more than usual interest, which may serve to justify a somewhat extended discussion of the individual pieces. . 1937B d an ; . wesFig1 A , t .22 section through north rampart, below flag footing (see also fig. 29, p. 324, below). Mr J. A. Stanfield reports as follows:— "The fragment (drawin mucs i ) gA he glaze.ha wornth d an , s disappeared fro e surfacemth n reliefi s . Nevertheles e desigth s s clearlni y recognisable,

1 PSAS, xxx. pp. 107-108. 2 JBS, xxv. pp. 196-197. 3 An analogy, from the earlier excavations at Birrens, is referred to in JBS, xxv. p. 199, where its implications are not realised. 4 Niederbieber, p. 19. EXCAVATIONS AT BIRRENS, 1936-1937. 311 e restoratioth d an n (drawin s certaini ) e design arcadeB ga Th .s e i n th , arches being double, connecte y astragalib d d supportean , pillary b d s with prong-like projections from the capitals, while each pillar is placed between a pair of wavy lines. The figure subjects are Bacchus (O. 566) and Diana (O. 109). A bowl by the same potter has been found at Chesterholm (AA4, xiii., , 12)8 fig. , but, sinc e connectioth e n between tha e Birrentth bowd san l fragment is not immediately obvious, a fragment from London is also illus-

] A5TANF1UP

. DecorateFig22 . d Samian from Birrens (A=B frod an m) London (C). (|.)

trated (drawing C), which bears the peculiar cruciform ornament of the Chesterholm bowl side by side with the Diana of the Birrens piece. These two motifs also occur together on fragments, not yet published, from Corbridge. As stated in the notes on the Chesterholm bowl, the style of these designs is characteristic of certain Central Gaulish potters whose manufactories were in all probability situated at Vichy, and whose most fruitful period of pro- duction occurred durin e principatth g f Trajano ee presenTh . t potter's designs contain ornamental elements that were used by several of the Vichy potters, notably RANTO and MEDETVS. Nevertheless, he cannot be identified with either of them, for the only signed example of his work known to the present writer bears the initial D. The cursive initial occurs, below decoratione th Cambridge pieca th n n fori f o ,e, o m37 e Museu Archaeologf mo y and Ethnology, on which the decoration is similar to that of the Birrens fragment, namely an arcade, whose double arches (one enclosing the same 312 PROCEEDINGS OF THE SOCIETY, MAY 9, 1938.

Diana) are connected by astragali and supported by identical pillars between wavy lines. Other figures on the Cambridge piece are Perseus (O. 233), a dancern interleave . a 281) (O d an , d snake-like ornamen e stylth en i to f RANTO. Additional figures, occurrin separata n go e fragment, undoubtedly part of the same bowl, are the group O. 238 and a dancer (O. 354). The work of this potter, who must now be referred to as D...... cannot be said to be uncommon, as examples of it have been found over a wide area that includes London, Colchester, Richborough, Verulamium, and Silchester e soutth Englandf n ho i , York, Corbridg Chesterhold ean northe th d mn i an , Leicester, Wroxeter Chested an , betweenn i r . remainw no t establiso I t s e connectiohth n betwee e wor nth f D..... o k . thad f RANTan o t MEDETVd Oan mean.followiny e Sb th f so g concordance: the signed bowls by the two last-mentioned potters being the well-known form 29 from Heddernheim stamped RANTOF, and the vessel of the same form at Kettering with the stamp MEDETI M. They will be referred to as the Heddernheim and Kettering bowls respectively. ovoloe Th (drawing Bclosels i ) y simila thoso t r fragmentn eo e stylth en i s of RANTO from London, Leicester, and Corbridge. The fine wavy lines (drawings B and C) are characteristic of Trajanic potters generally and, -wit e seven-bead,hth rosette (drawine ) occuth C g n o r Heddernheim and Kettering bowls. The Bacchus (drawing B) occurs on a fragment of form 37 from Leicester, a fragmen e n Britiso th d t a htan Museu . 1389m(M ) wit e samhth e ovolo, both in the style of RANTO; and also on a form 30 at the Guildhall Museum, London conjunction i , n wit e peltahth that e Heddernheioccurth n o s d man Kettering bowls. The Diana (drawings B and C), although it has not been met with by the present writer on any bowl directly in the style of either RANTO or MEDETVS, occurs (drawing C) • in conjunction with the little butterfly-like ornament used on the Heddernheim and Kettering bowls. Both the Diana and the "butterfly" occur on the Corbridge pieces already referred to. The double acanthus leaf, (drawing C), which also occurs between the arches Cambridge onth e bowl, also occurKetterine th n so g bowl. .Another elemen e f Heddernheidesigo tth n no m bowl which, thougt hno present on drawings B and C, occurs on fragments in the style of D...... is a beaded ring simila o thost r e use IOENALIy b d d DONNAVCVSSan t bu , like RANTO's variety of that ring in having a tiny five-pointed star within the ring. This beaded ring appears on a fragment of form 30 in the Guildhall Museum, which also shows the pillar, the double arch, and the ovolo of drawing B. There is therefore good evidence that the potter of the Birrens fragment, the initial letter of whose name was D, was a Trajanic potter, perhaps directly associated with MEDETVS and RANTO. The date of the Birrens piece woul rathee db r later tha e Heddernheinth Ketterind man g bowls circay sa , A.D. 110-115." The date assigne thio dt s piec generan eo l groundcheckee b y dma s by reference to site-evidence from the north of England. Both at Corbridg Chesterhold ean same wore mth th f eko potter occurs associan i , - EXCAVATION BIRRENST A S 3 , 31 1936-1937.

tion with deposits which start in the Flavian period, and are sharply interrupte t aboua d e pointh t t wher e depositth e Hadrian'n o s s Wall begin t Higa ; h Shield, e otheth n r o 1hand pieca , e same byth e potter has been foun association di n with pottery clearly starting with Hadrian. Thao say t have s i w ,t a e"borderline " potter befor , whosus e e work ma expectede yb Trajanin Hadrianin o o d an conls i cy t y i b sites d an ; referenc associatee th o et d finds tha t wil i tpossibl e b l judgo et sitee eth - dating. Fortunatel t Birrena y e associateth s d finds, thougn i w fe h number, hav ecleaa r stor tello y characteristie t plac n i th ; f eo c fabrics e Flavian-Trajanioth f c period e deposith , t from whic e presenth h t piece came included two examples of the black-fumed cooking-pot, e nortwhicth f Britain o hi h s typicai ne perio th f o dl from Hadrian onwards. Ther thus ei necessito n s supposo yt e occupatio t Birrenna s under Trajan; but the occurrence of this fragment, and some other pieces discussed below (p. 322), provides satisfactory evidence of occupa- tion starting at about the same time as in the forts on Hadrian's Wall. 2. Fig . 193723 . ; underlying north rampart, eas. tA section . J r M . Stanfield supplie followine sth g note:— "A small fragment, fortunately large enough to show the form of the vessel of which it was a part, namely the hybrid form 29/37. The central moulding bears a string of conjoined astragali bordered by rows of small, neat beads. Ove e upper th this n ro , frieze a serie s i f acanthu, o s s calices placed sid sideeby , with anothe beadof s rrow above and, finally remainthe , s of a series of festoons with pendants that terminated in sharp-pointed leaves. Less remain lowee th f rso frieze t ther indicationbu e , ear s tha decoratione th t consisted of medallions (of the same type as the festoons) alternating with some other ornament. In fig. 23, A is a drawing of the fragment itself, and alss i t oi show section ni e restore s parth na f o t d profile, base bowa n f do o l similar shape from Heiligenberg (Knorr, Rottenburg, pi. viii. 7) . As concerns the style of decoration, it is true that strings of conjoined astragali were much used on pottery from Trier, notably by the potter Dexter (Folzer, pi. xv. 14), and also festoons somewhat similar in character to those of this piece (ibid., pi. xv. 13, 18, and 19); and conjoined astragali were also used on La Madeleine ware (Ricken, Saalburg, pi. ix. 19, and x. 1). On the other hand, althoug n acanthuha s caly s alsoxwa use t Triea d r (Folzer, pi. xxxii. 900), as used there it has not the drooping ends of those on the Birrens fragment; these are much more like Ludowici's type O. 1, used by seven Rheinzabern potters, including lanus and Reginus, who both appear to have worked at Heiligenberg before moving to Rheinzabern (Oswald, Index of Potters' Stamps, voce).sub Apart froe ornamentamth l types mentioned above, however, thers i e little resemblanc Triero et Madeleinea L , r Rheinzabero , n presen e warth n ei t fragment, and much closer parallels exist in certain pieces, from a pottery

1 . 349 AAp . .4xv , 314 PROCEEDING E SOCIETY , 19389 TH Y F O S. MA ,

which Bicken doe particulat assigy sno an o nt r placeSaalbure th n ,i g collection (Ricken Ricken'n ,occur 7 Saalburg,O d same . an s th 9) s6 ed an pi , 7 . , xiv6 , 5 . acanthus calyx with drooping ends, a usesimila n i d r way, namely, side hy side—in 7 taking the place of an ovolo, and in 6 at the base of the decoration. Furthermore, the small triple ornament in the medallion of the Birrens sherd occurs on all four of the Saalburg fragments, on which rows of small beads alse ar o seen thao s , t ther littls ei e doubt tha piecee th t s from Birrens and,the Saalburg are the products of the same pottery.''

Fig . Eas23 . t Gaulish bowl, form 29/37. (J.) 3. Fig. 24, 1. 1937; Site XXII East, unstratified. Mr J. A. Stan- field writes:— "Thi Madeleinpara s bowsa i L f o tn i l e fabric, wit hvera y deep plain band between the lip and the decoration. All the types are shown by both Folzer Ricked Madeleina an L s na e types, viz.: Folzer, pi. xxv. Ricken, pi. vii. Acanthus calyx 74 25 Astragalus 100 8 Triple leaf 92 14 Ovolo 119 C Rosette . 108 1 A similar pane f superimposeo l d acanthus calices occur fragmena n o s t from Friedberg (Ricken, pi. x. 10), and astragali placed athwart bead-rows are commo thin no s ware , (c/etc.14 . , ;Ricken 6 Folzer, 3 , 1 , , ,. pi 29 piix .. i . 39, etc.)." EXCAVATION BIRRENST A S , 1936-1937. 315

Pig. Decorate24 . d Samian ware. (J.) 316 PROCEEDINGS OF THE SOCIETY, MAY 9, 1938.

The products of La Madeleine are seldom represented on sites in the north of Britain I hav ;t note otheey no dan r examples from Scotland, but there are several vessels included in the Corbridge collection; none of the Corbridge pieces is close enough in style to the present bowl to •warrant a detailed comparison here. 4. Fig. 24, 2. 1936; unstratified. Mr Stanfield writes:— "This piece, in Lezoux ware, is unmistakably the work of the potter DIVIXTVS e seateTh . d figur f Fortuneo e . 801(O ) occur n for o sfro7 m3 m Silchester, form 37 from Corbridge (AA3, viii. p. 191, fig. 13), form 30 from Caerleon (Lee, Isca Silurum, pi. xii. 3), forms 30 and 37 at Colchester, and anothen o r for fro7 m3 m Corbridge (no t published)tye l stampeal , d DIVI. F X • The caryatid (O. 1207 A) occurs on at least eight bowls bearing the same manstampn o d y an more,same th en i , style whicn o , stamo hn preserveds pi ; it is perhaps the commonest figure-type used by DIVIXTVS. The third figure whicf o , h onl ylege par th preserveds si f o t seatee th s i , d Bacchus 571. (O ) which occure bowlth n so s from Silchester, Caerleon d Colchestean , r referred aboveto anythinIf . g further were neede provdto attributioethe the nof piece, it would be the ring-terminal of the bead-rows, which occurs on as man seventees ya n bowls stampe thiy db s potter." 5. Fig. 24. 19373 , ; Site XXII e Westsamth en i , occupation-layer as fig. 31, Nos. 1—8, below. A small, worn fragment from a bowl by the "Potter of the small S" (cf. CW2, xxxvi. p. 136), whose products are widely represente Antoninn do e sites r examplfo , t Balmuildyea , Mum- rillsd Newsteadan , panthere Th . 1570). (O , whose head just shown o s e presenth t f thio fragment se potter'on s i , s commonest typess hi ; commonest ovolo is that in which the tongue is replaced by a knob projecting from the outer line of the egg, at the right-hand side, as on Mumrills, Nos. 48 and 54; the former piece also shows the leaves in e fiel whicf th do pottee hth extremels rwa y fond. 6. . 1937Fig4 , .24 ; Site XXII e Westsamth en i , occupation-layer. A large piece, partly burnt, from a bowl decorated in free style; it comes from a rather worn mould, so that the decoration has lost much of its detail figure-typee Th . lion1450)a . e (O sar , whose tail just showe th t sa s forepar . a 1772)stage righthi th (O d e typed leftt th an e ,a an ; t ar s assigned to various potters by Dr Oswald, but the only two shown to have used both of them are ATTIANVS (who stamps OF ATT) and CRICLRO presene th ; t piece migh assignee b t formere th use o o dt swh , similar conventional herbage, but the ovolo looks unlike any known to have been use Attianusy db . 7. Fig. 24, 5. 1937;".Site XXII West, overlying the filled-in ditch d sealean latey db r walls (cf. . 325 p fig , ,.30 below). Para bowl f o t , form 30, in the style of CINNAMVS. The figure-types are Venus (O. 331) EXCAVATIONS AT BIRRENS, 1936-1937. 317

and a warrior (O. 204), both of which occur commonly on work stamped by this potter; Mumrills, No. 32, is a larger piece from a bowl probably made in the same mould. 8 .. 1937Fig6 , 24 .; unstratifled Stanfielr M . d sxipplie followine sth g description:— "Form 37 in rouletted technique, from an East Gaulish pottery. Three deep grooves separate the plain zone from the rouletted decoration, which laterain si l bands placed close togethe sometimed an r s overlapping." 9. . 1936Fig1 , .25 ; west gate, unstratified. Parrathea f o t r small bowl of form 37; Trier fabric. The ovolo, Folzer's type 944, is not very

Fig. Trie25 . r ware fragments , Birrens1 : , Corbridge2 ; , Housesteads3 ; . (J.)

clear, partly because the bowl has been made in a worn mould, and partly through defacement sustained when the rim was being attached. figure-typeThe boxerstwo are s(Folze r 524, 525), which occur frequently on vessels with this ovolo; the name of the potter who used it is not known t Folzerbu , , discussin work,s ghi e e closth th 1f assigneo o t m shi second centurbeginnine th thirdd e yth an f gdatino a , g supportee th y db occurrenc s producthi f eo t Niederbieber.sa s wor Hi seldos k2i m found e nortith n f Britaino h e hav t w note ;no ey othe an d r examplen i s Scotland, and there is only one piece showing the same ovolo in the collection at Corbridge. In order to give a better idea of his style of decoration, we illustrate the Corbridge fragment and four pieces of a

1 Op. cit., p. 79 f. 2 Niederbieber, pi. vii. 34 and 35. 318 PROCEEDINGS OF THE SOCIETY, MAY 9, 1938.

bowl from Housesteads: , Corbridge;2 Fig , 25 . find-spot unrecorded. Ovolo befores lefo a t g t (Folzedo ; r 651); tairigho t g l tonldo (Folzef yo r 642). , Housesteads;3 Fig , 25 . underlying buildin ge vicus. VIIth n i I Ovolo s beforea ; lionboard an (Folzer 593, 610); lion (Folzer 589); hind leg (Folzeg s do onl f o yr 650) d treesan , (Folzer 737, 772)e loweTh . r e forwreatth mf o particularls hi y characteristi f Trieco r productsd an , another typical feature is the use of plain guide-lines instead of bead- rows. The glaze of this bowl is of the clear orange tint which Folzer notes as a common feature of the potter's work.

B. Other Pottery. In considering the pottery other than decorated Samian ware, it seems necessary to begin by setting forth the principles on which its value for dating purposes should be estimated. There are several points to bear in mind.- In the first place, not all types had a sufficiently restricted greaf o e tb lif valuo et thin ei s connection; thus flat-rimmee th , d platter, in the black fumed ware which first appears in Hadrianic deposits in the north of Britain, persists well into the third century, with little dateable variety in its profile ;and it is often difficult to attempt a close dating

of cooking-pot e samth sn e i 2 fabric1 . Though stratified example f sucso h vessel e illustratear s t beedno ns below thoughha t i , t worth whilo t e quote extensive parallels from other sites. And when parallels are quoted, to be of real value they must only be taken from securely stratified deposits, preferably dateable within exact limits thar fo ;t reason, unless shoo t s wi t thai typa t e occur Antoninn o s e s seldo sitesha t mi , been necessar o quotyt e parallels from Newstea e r fro o th de fort mth n o s Wal Piusf o l ; quite apart fro uncertainte m th terminae th o t s yla date of their occupation, e scarcit3th f stratifieyo d pieces assignablr o e on o et othe theif ro r successive periods necessitates looking elsewher parallelsr efo . Fortunately a valuabl, e serie f materiao s s availabli l e from Hadrian's Wall, particularly fro e mile-castlemth d turretan s s excavater ou y b d Honorary Fellow, Mr F. G. Simpson, whose reports * provide the essential starting-point for a study such as this. There will be frequent occasion to refer to the periods of Hadrian's Wall, so that it may be desirable to insert her eschedula themf eo : 5

1 Of. Birdoswald, fig. 16, 73. 1 For the definition of cooking-pots, as opposed to jars, cf. CW2, xi. p. 450. 3 See "below, p. 343. 4 CW2, xi. pp. 390-461; CW2, xiii. pp. 297-397. 5 Cf. AA4, vii. pp.'164-74; AA4, xv. p. 267. EXCAVATIONS AT BIRRENS, 1936-1937. 319

Period I A: circa A.D. 122-162. Perio : circaB dI A.D. 162-196. Period II: circa A.D. 205-297.

Period III: circa A.D. 300-367. Period IV: circa A.D. 368-383 or later.1 Periods I A, I B and II are of obvious importance for a study of the pottery from Birrens; types which occu Perion i r datee b I onldI n dyca securel thire th do y t century demonstratd an , e thafore th tt continued e occupieb o t n thai d t e knowperiodth d nan ; re-buildin f Birrengo s in A.D. 158, coming so close to the end of I A, invites a comparison betwee second-centure nth y levels her thosd eHadrian'an n eo s Wallt Bu . of tha te sequelmorth e musn i ew ; t r examinatioreturou o t n f o n principles. The next establishee poinb o t te definitio th s di a usefu f o n l type. By that is implied more than the occurrence of pieces similar in profile d fabrican ; wha s mori t e importan s consisteni t t stratificationn I . other words, an isolated piece, even though securely stratified, is not of freaka e b grea ;y 2 tma similavalue t i r fo r, pieces coming from deposits of different periods (unless, for reasons discussed below, they can be showf thosintrusive o b e eo nt on periods n ei ) cannot provide evidence for clos r moro e o edating tw vessel e samt th bu f ;e so profil fabricd ean , coming from deposits of the same period, allow the confident dating of unstratified parallels. It is perhaps necessary to emphasize the import- anc fabricf eo , sinc n examinatioea e pieceth f no s themselve needes si d for its identity to be established, whereas similarity of profile can be shown by drawings; and experience shows that similarity of profile alone is not necessarily a safe guide.3 At this stage it seems necessary to insert a caution as to the interpreta- tion of stratification. The mere occurrence of a piece in a stratified deposit doe t necessarilno s y mean tha e vesseth t l from whic t hcomei s was in use during the period when that deposit was formed. We may exclud e possibiliteth introductiof yo n fro mlatea r leve subsequeny b l t disturbance, for that will mean that the deposit is not completely stratified. But it often happens that pieces lying about a site long occupied will find theiint occupation-layeoan way r late far rdat in r e tha perionthe d 1 Cf. Hedley, The last days of Corstopitum and the Roman Wall—the coin evidence, in AA4, xiv. pp. 95-102. 2 E.g. Appletree turret, pi. xvii. 63, unjustifiably treated as a type by me in AA4, vii. p. 173, and by Collingwood, Archaeology of Roman Britain, p. 222. mortariume th . Cf 3 , Birdoswald, s profilit s suggeste; ha e 10 fig , .13 earln da y second-century date to competent judges, but its fabric is that commonly found in hammer-head mortaria, and it belongs to the close of Period II. 320 PROCEEDINGS OF THE SOCIETY, MAY 9, 1938.

of their manufacture breakaged an e us , ; 1 such piece usualln ca s e yb distinguished without difficulty e studenth y f b Romao ,t n ceramics, fro e materiamth l contemporary wit e occupation-layerth h s i t i d an , necessary to describe them'as "strays," in contrast to the "survivals" or pieces characteristi e perioth f co d immediately preceding onld an ,y rarely represente e perioth questionn n di i t woulI .t havdno e been necessary to stress this point if the distinction between strays and survivals had been observed more closely by previous writers. And

there is another point which seems worth stressing, namely, tha2 t it is sometimes possibl arrivo et closea t ea r dating than merel givea o yt n period, by observing the circumstances in Avhich a particular vessel was discovered. Thusclaa n yo , floor, whic mads hseverawa p eu l timen si course th perioda f e o piecee th , s trodden deepest int clae likele oth y ar y belono earle t th y o g t periode year th t f pieceo sbu ; s lying amone gth debri f destructiono s y contrastb , , wil s close.lit hav t a ee beeus n ni

perhapAns i t di clearese sth t indication that vessel3 s belonvere th yo gt a period f o f thed i , e founen yar d complet r nearlo e y completn a n i e occupation e normalayerth r ;fo l4 practic carro t s y ewa rubbis h away and tip it outside the fort; hence the scarcity of pottery in barracks of e secon th third an dd centuries, with which northern excavators have long been familiar. placf o t observo et ou e b emistak a t thas i no Finally t i ty o et ma t i , loo parallelr r afieldkfo fa o , s to dealine unlesar e sw g wit producte hth s of a great exporting centre, such as Samian or Castor ware, or unless we e confrontear d with vessels which ther s reasoei regaro nt s unusuada l importations into the district with which we are concerned; it is true that the interaction of Italic and La Tene elements produced a somewhat similar serie developmentf so potterye th n si fashion Westerf so n Europe,

t therbu sufficiens ei t evidenc shoo et w tha incidence tth thosf eo e develop- 5 ments varied considerabl n differenyi t districts, even withi e samnth e province.6

1 It is hardly necessary to quote instances, but cf. Corbridge, 1938, fig. 8, 14 and fig. 13, 6, pre- Hadrianic stray Antoninn a n si e deposit. 2 Thus, at Mumrills a typically Flavian carinated bowl with reeded rim (fig. 101, 1) occurred in Level B of the westmost Antonine ditch, and was interpreted as a survival into the Antonine fragmena periods wa f s "rustico a ,t " ware (fig. 103 ) foun4 , d "a littlbeneaty ewa modere hth n surface." 3 This point was first made, by implication, in the report on Poltross Burn milecastle, p. 447 et seq., where "early first period" material is distinguished from "first period (mostly in debris)." Cf. Birdoswald, figs. 13, 8; 15, 42 and 44—from the close of Period II; and Bitterling's 4 observations in Hofheim, p. 85. S. N. Miller in Balmuildy, p. 82.

Cfe cas th .black-fumef eo d ware, referre alreado dt occurrins ya depositn gi 5 s fro time mth f eo Hadria6 n onwards in the north of Britain (cf. AA4, xv. p. 229); in'Wales it began to appear before firse close th th tf ecenturyo . EXCAVATION BIRRENST A S 1 , 32 1936-1937.

Judged in accordance with the above principles, the evidence pro- vided by the pottery from Birrens is remarkably clear. Much of it, indeed, for the reasons outlined above, cannot be dated very closely; denudatioe th d uppee an th f n o rdifficulte levelsth d accesf an ,e yo th o st earliest strata, combine to make the series as a whole rather scantier at the beginning and the end of the occupation than could have been instructivs i t wishedi t Bu noto et .correlatione eth s whicbees ha n t hi possible to make between Levels II and III at Birrens and I A and I B on Hadrian's Wall; the jug and mortarium l belonging to the close of the period represented by Level III, the former underlying a wall which n 189i d bee5ha n e take"primary,b o t n " justif e attributioth y f o n Level IV to the Severan reconstruction; and there is ample variety of types whic Hadrian'n ho s Wall occur onl Perion yi . dII Som thesf eo e types can be assigned to the end of that period, particularly the haematite- washed mortaria represented most strikingly at Birrens by the piece found underlyin water-channele th e sid f gth o e , nea norte th r h gate;2 so that there can no longer be any question but that Birrens, like the other outliers of Hadrian's Wall, continued in occupation throughout the third century unfortunats i t I . e tha latese th t t leve produces ha l d so little pottery, so that we are unable to estimate how long into the fourth century it lasted; but the fine jar illustrated as fig. 35 is a handsome additioe growinth o t ng corpu f earlo s y fourth-century types. Fig. 26. 1936; stratified pieces from Site XIX.

Pig. 26. Stratifled pieces from Site XIX. (J.) 1. Leve . I l Flange fragment fro mSamiaa n bow f Curie'o l s type 11; on the upper edge of the flange is barbotine decoration, as normally on this type. s Curidescribei 1 1 e Oswaly db Pryced dan , Terra Sigillata, . 212p s especialla , y characteristi e Flaviath f co n period t persistinbu , g int time oTrajanth f lowes eo it t carriere bu b limi; n dca t rather farther, for an example occurred on Hadrian's Wall at Birdoswald turret (p. 349). As far as the present piece is concerned, it seems to come late in the developmen shar e typee th th r pf fo ,o t downwar d flange turth f no e seems 1 Pig . 27 . . 2 14 Fig , 31 . 322 PROCEEDINGS OF THE SOCIETY, MAY 9, 1938.

without parallel on any Flavian site. It must be left undecided whether we have her n exceptiona a a elat r o e l vessel t i canno: takee b t s na necessarily pre-Hadrianic. 2. Level II. Deep, flat-rimmed platter, with slight chamfer between walle th e basd ;th ean blac k fumed ware. Fro time m th Hadrianf eo , when it first appears in the north of Britain, there is little variation in the rim-section of this type; analogies could be cited from both second- century periods on Hadrian's Wall. But the fabric of this vessel, a rich black, wit e surfachth e highly burnished s besi , t parallele depositn di s of period I A. similaa f o 3rm . platterRi Leve . II l ; onc badlew blackno y t bu , burnt. 4. In debris overlying the flagged floor of Level IV. Outbent rim- fragmen grea f o yt fumed cooking-pot . Throp,Cf . 2 2 pi .d xxvian 1 2 . (associated wit ha hammer-hea d mortariu whitmn i e pipeclay waref o , the type that first appears on Hadrian's Wall circa A.D. 300; many more parallel same th f eso date coul cited).e db 1 . 19361 , Fig;27 . WestSitX eXI , mostly underlyin norte gth h walf lo greatee Th witLeveg . rju h para IV l pinche f o t d neck hardn i , , rather sandy, grey ware; jars in late second and early third-century deposits on Hadrian's Wall are often of a similar texture. Jugs with, pinched necks occur sporadicall l periodal n i y s fro e firsmth t centur o tht y e fourth (indeed, they long outlast the Roman period), and it is rather from the fabric and the contour of the body of the jug that an estimate of dating derivede b i o st thin I .s cas s beefabrice eth ha n s a said, affinities ha , s •with that common circa A.D. 200, and the rather bulbous body best suits same th e period. . 19362 , Fig;27 . yardfounw fe se precedinda easth f o t g vesseld an , at the .same level, though not underlying the north wall of XIX West. Most of a mortarium, in fairly hard, reddish ware; the grit is fine and mainly white closA .occur em paralleri t Poltrosse sa th o lt Burn,7 pi. .iv (first period, "mostly in debris," i.e. latter part of the second century), a piece with an illegible name-stamp on it; I have not had an oppor- tunity of examining this piece recently, to see if the stamp can now be deciphered. There is a somewhat similar rim from Throp, pi. xxvi. 2; Corbridge, 1911, 45 is in a wholly different fabric, and should not be confused with this type lightee th ; r Hadrianic rims, Birdoswald, fig, 13 . provid instructivn ea e contrast.

site-evidenc e vien th I 1 f wo e fro norte mth Britainf ho t seem,i s probable tha three tth e intrusive rims in the alley at Birdoswald (p. 191, h-k) belong to the Constantian rather than the Severan reconstruction e Constantiath ; n builders carried their-foundations deep down (Birdoswald, . 171)p . EXCAVATION BIRRENST A S , 1936-1937. 323

F.c. ,937

. FigStratifie27 . d vessels from Site XIX. Fig. 28. 1936; Site VIII, Level III +. 1. Rim-fragment from a Samian cup; I have not been able to find a parallel to this form. 2. Wall of a Samian cup, form 33 ; outside, on the lower part, is the name ownerth f eo , Gaiubelow)4 33 . s p (fig., 36 . 3. Black fumed cooking-pot. 4. Cooking-pot, once fumed; the surface is badly burnt. 5. Fumed beaker. 6. Grey jar . Corbridge,cf ; 1938 Antoninn (a 0 , 2 fig , 7 .e piece). 7. Grey fumed cooking-pot; cf. Poltross Burn milecastle, pi. iv. 24 (Period II). 8. Black fumed platter. 9. Similar platter. 10. Dark grey roll-rim platter, fumed but not burnished. 11. Platte lighn i tm rri buf f ware. 12. Lipless platte blacn i r k fumed ware, with lattice e scoreth n o d outside. VOL. LXXII. 22 324 PROCEEDING E SOCIETY , 19389 TH F Y O S. MA , 13. Fumed platter, burnt yellow; cf. Corbridge, 1938, fig. 8, 5 (second Antonine level). From the same deposit came part of a "hunt cup" in blackish Castor ware, and much of a Samian platter, form 18/31, with the incomplete stamrp AE [ (fig. w36, p. 334, below); as often happened in a wood fire, the ^ET* T: £6.

Y 12 I I 1 \^r IO r r . FigRim28 . s from Site VIII, . Leve(J.+ I ) II l latter vessel has been burnt black (cf. Newstead, p. 230, a platter by Cintugenus, similarly burnt). . Fig193729 . ; below nagged foundatio f norto n h rampart, west section. 1. Black fumed cooking-pot. For the rim-section, compare Bird- oswald turret, pi. xvi. 11 (period I); High House turret, pi. xvi. 39 (period I A); Chapel House milecastle, pi. liii. 56 (period I B); Balmuildy,

ih '37 Pig. 29. Deposit below the existing north rampart, pi t doe I .t see xlv . sno m 14 . possibl givo et eclosea r datintype th eo gt than Antonine. 2. Black fumed cooking-pot. The closest parallels that have been noted are Chapel House milecastle, pi. liii. 59 (period I B) and Balmuildy, pi. xlv. 13 . basd an e fragmentm 3Ri . moderatela n si r froja ma y hard, reddish •ware I hav t .com no e e acros closy an se e paralle th forme t th bu o , t l obvious ha m ri s affinities wit carinatee hth d bowls, whose rim longeo sn r sho reedingy wan , that occu Hadrianin ri c deposit Walle th n .so 4. Plattereddisn i m ri rh ware with buff-washed surface rime Th -. EXCAVATIONS AT BIRRENS, 1936-1937. 325

type does not seem to occur elsewhere in the north of Britain; the fabric suggest Hadrianisa c date. Fig. 30. 1937; XXII West, overlying the filled-in ditch and sealed by later walls. 1. Samian cup, form 33, stamped GNATI-M. For a note on this Antonine potter see p. 334, below. 2. Wall fragment fro msimilaa r cup. 3. Rim and wall of a jar in fairly hard, blue-grey ware, grey in fracture. Contrast the rather soft fabric of the Hadrianic jars at Birdoswald (p. 191); rime th . approache highe sth , third-century type. 4. Cooking-pot rim. 5. Beaker rim in black fumed ware; a wall-fragment, from the same deposit, has a handle (cf. Old Kilpatrick, p. 46; Newstead, p. 256; Bird-

. PigSeale30 . d deposit from Site XXII West. (J.)

oswald, p. 192). For the rim-section, cf. Appletree turret, pi. xvii. 88 (period IB); Birdoswald, p. 194. 6. Platter rim, of the well-known Antonine type, Newstead, pi. xlviii. 42. fragmenm 7Ri . t fro mlarga e platte coarsen ri , sandy, dark grey ware. 8. Roll-rim platte fumen ri d ware. Fig. 31. Stratified pieces from various deposits examined in 1937. Nos com8 1- . e fro burne mth t laye Sitn ro e XXII West (cf. . 299p , above). . Flange1 d bowl harn ,i d reddish-buff ware . Caerleon,cf ; Archceologia Cambrensis, (date7 19295 . d No , " Hadrian-Antonine "). 2. Black fumed platter-rim; cf. Appletree turret, pi. xvii. 67 (period

3. Disk-mouthed jug-neck n browi , n ware with polished surface. The type is a long-lived one; cf. Newstead, fig. 33, 11 and p. 263; Colchester Museum Report, 1930, No. 146.30, p. 35; Templebrough, pi. xxxiii. B, 223 . 11 p ver a 5( d yan close parallel). 4. Black fumed platter rim. 5. Worn fragmen flat-rimmea f o t d platter. 6. Worn rim-fragment of a fumed cooking-pot. 7. Reddish-buff jar rim, in coarse soft ware. 326 PROCEEDING E SOCIETY , 19389 TH Y F O S. MA ,

8. Black fumed cooking-pot rim; cf. Chapel House milecastle, pi. liii. 53 (perio ; Corbridge,B) dI 1938, (secon2 fig , 9 . d Antonine period). 9. Black layer overlying roadway, XXII East. Black fumed cooking- pot; cf. Birdoswald 42 (period II) for the rim-type, though that is a different type of pot. 10. North rampart, west section; in the bottom of the turf revetment V r

r 7

12

. FigStratifie31 . d pieces from various deposits. (J.) at the front, beyond the stone kerbing. Rim fragment of a Samian platter, Curie's type 23 (an Aiitonine type). 11. XXII West, in the body of a buttress supporting the late east- west "wall. Fumed flat-rim platter. 12. As No. 11. Grey-black fumed cooking-pot. 13. XXII East, underlying the secondary hearth (cf. fig. 15, above). Much of a grey fumed platter, of the deep chamfered type with flat rim, Birdoswald 65 and 66; the rim of the present example is a good deal coarse rBirdoswale thath n no d bowls. . Nort14 f XXIho I East , e clasealeth y y layeb d whicn o r ha sid e water-channee stonth f eo l rested. Rim-fragmen small'mortariua f o t m in hard, red-brown ware, with a fine heematite slip on the rim; white EXCAVATION BIRRENST A S , 1936-1987. 327

and brown grit. The type is well attested in third-century deposits on Hadrian's Wall . Poltrosscf ; Burn milecastle, Highd an 0 1 House . piiv . milecastle, pi. xviii. 113 , thoug14 , e sam. botth h No e n hi r fabriou s a c belongin largeo gt mord ran e open vessels; another exampl same th f eeo form, thoug a differen n hi t fabric s beeha , n foun t Winshielda d s mile- castle (unpublished), as Mr F. Gr. Simpson has been good enough to inform me threr o . o eTher vesseltw same e th ear f eso type included amone gth material foun t Birrena d preservew 189n e Nationano i s th d 5n an di l Museum of Antiquities (NMA, FP 108, 119), and another, found on Site VIII in 1936, is figured below, fig. 33, 1. Fig. Potter32 . y from various sites.

rr 12

Fig. 32. Pottery from various sites. . 19371 ; unstratified, fro r earlmfo easyt nortcu t h ditch largA . e piece from a cooking-pot with high, rather outbent rim. The exterior of stilm ri l showe th lightle sth y scored wavy line, whic usualls hi y regarded as an indication of second-century date; but an example has occurred, on a somewhat similar rim, in an early third-century deposit at Denton Hall turret (pi. li. 13), and another in a third-century deposit at Corbridge (Corbridge, 1938 e hatchin, fig th e vessele bod , 14)th d .7 th f an yn ;o g o , at an oblique angle to the horizontal, and the horizontal scored line above the hatching, are normally met with on cooking-pots of the latter part of the third century and the first hah0 of the fourth. This vessel cannot wel earliee b l r tha thire nth d century. 2. 1937; Site XXII West, in the same layer as fig. 31, 1-8. Rim oa cooking-pof t belongin e samth eo g t genera e lprecedin typth s a e g 328 PROCEEDING E SOCIETY , 19389 TH F Y .O S MA ,

vessel; abou piece0 similaa 15 tf thi d o st wer san po r e found hereo to , badly burnt and decayed for reconstruction to be practicable. beakea f o 3blacn .m ri 1937Ri k . fume2 ; . origidNo wares na e on ; piec bees eha n badly burnt, anothe unburnts ri . harde th jara n ,i f ,o blue-gre 4m . 1937Ri . y2 ; war . origiNo e s na characteristi f third-centuro c y deposit Hadrian'n o s s Wall (cf. Poltross Burn milecastle, e sharpl th . 451) p r fo y; outbent rim. Dentoncf , Hall turret, pi. li. 15, and Corbridge, 1938, fig. 7, 3 (these two vessels are both cooking-pots, and both come from third-century deposits). 5. 1936; unstratified. High-rimmed cooking-pot; cf. Birdoswald turret, pi. xvi. 21 (Period II); . High House milecastle, pi. xviii. 120 (Period II). 6. 1936; unstratified. Grey jar, with outbent rim; cf. Poltross Burn milecastle, (Perio4 2 . piiv .d II). 7. 1937; Site XXII East, in or over the black occupation-layer overlying the roadway. Cooking-pot rim, of the same high type as No. 2 above, but lacking the wavy line. 8. 1937 . 7 Cooking-po; . origiNo s a n t . rimAppletreecf ; turret, pi. xvii. 83 (Period I B). 9. 1937; Origin as No. 7. Platter rim. . 193710 . ; topsoi section i l n through east rampart. Beaker rimn i , friable grey clay, probably burnt Birdoswald,. Cf . (Perio2 4 , fig15 .d II). 11. 1937; origin as No. 10. Roll-rim platter. 12. 1937; topsoil in section north of XXII West. Outbent cooking- pot rim. 13. 1937; origin as No. 12. Flat-rim platter. 14. 1937; origin as No. 12. Platter rim. Fig. 33. Mortaria from various deposits. 1. 1936; Site VIII, Level II . IAbou+ t hal mortariua f hardmn i , reddish-brown ware, with traces of haematite slip on the rim; the interior is thickly sprinkled with a crystalline grit; the spout has been broken off. This is a larger example of the third-century type discussed under fig. 31, 14 above. 2. 1936; Site VIII, Level III+. Mortarium in red-brown ware, once wit rime hdara th sli d ;n kpre o spars e whit browd ean n gritA . somewhat similar rim occurred in Period-I "mostly in debris" at Poltross Burn milecastle. 6) (pi . .iv 3. 1936; Site VIII, Leve I +II l. Light brown ware, with tracea f so cream wash, and a name stamp, impressed twice, which I have been unable to decipher. 4. 1936; unstratified. Rim-fragment in rather rough, yellowish-drab EXCAVATION BIRRENST A S , 1936-1937. 329

ware withou y slip-coatingan t e fabri d rim-sectioTh an .c n together leave no doubt that this comes from a mortarium by the potter ANANVS or ANAVS, one of the Antonine potters whose work is most widely distributed in the north of Britain (cf. Cartridge, 1938, p. 276).

(fill's

Fig. 33. Mortaria. (J.)

5. 1937; Site XXII West, unstratified. Much of a mortarium in red ware, wit hthina , dar glazed kre e spou th ; smals i tneat d d an l an , there are two lugs or handles. There are one or two examples of the same type in the Corbridge collection; cf. Wroxeter, 1912, types 74, 78, where it is suggested that the type was introduced into Britain about the end of the first century, and lasted well into the second, "but their disappearance is not yet dated with any accuracy"; it is extremely uncommon in the north of Britain, and I cannot quote a stratified example; but there are 330 PROCEEDINGS OF THE SOCIETY, MAY 9, 1938.

tw r threo e represente e materiath n di l fro e excavationmth f 189o s 5 (NMA, FP 111, 114117)d ,an . 6. 1937; Site XXII West, unstratified. Hard yellowish buff ware, with brown grit potter'a ; s stam juss pi t amissin eithen go e r th sid f eo spout vessee th bees ; ha l n overbaked. 7. 1937; Site XXII East, unstratified. Hardish red ware, "with haematite wash; cf. Poltross Burn, pi. iv. 8 (Period II). 8. 1937; north gate, unstratified. White pipeclay ware, with pink grit thickly sprinkled on a fragment from its base (not drawn). . The fabric and grit recur on some examples of the well-known Constantian typ hammerheaf eo d mortarium presene th hardlt m bu ri ,t y look lats sa e as that period. 9. 1937; XXII Westsame th en i , deposi piecee th s sa t illustraten di fig. 31, 1-8. Whitish buff ware, with part of the retrograde stamp of Mossius; there is a more complete example of the same stamp from Rough Castle (NMA, FR 341), and the potter's name is established by stamps from Lincoln (British Museum, cf. B.M. Cat., M 2793; I have examine e piecdth e myself d Wroxeteran ) (1912 . 10)No , . Fig. 34. Four mortarium rims with makers' stamps, from earlier excavation t a Birrenss e preservear e , Nationath n i d l Museuf mo Antiquities, and are illustrated in this figure. 1. FP, without a number; stamped AVSTiNi MANu. Stamps of Austinus, fro t mleasa t three different dies, occu t Amblesidea r (CW. 2,xv , whers bee56 ha n. t ei p misread), Carlisle (CW , xvii., pi. xvii , Cor3) . - 2 bridge and Chesters (unpublished) in the north of England; and at Newstead (unpublished; NMA o number)n , , Mumrillsr Ba , 1) (fig , 94 . Hill . 70)(p , Camelod Balmuildyan ) 18 n d (NMA an (pi 2317 X .1 F ,xl, )B . in Scotland. This distribution make t probablsi e tha pottee tth r worked in the north of Britain, but there is at present no evidence sufficient to indicate the place where he worked. P 1942F . , PSAS, xxx . .186p ; stamped ]NANIt ye I hav t . no e me parallea t thio lt s stamp. 3. FP 193, PSAS, xxx. p. 186, 2; stamped SARR in large letters. Sarrus used a large number of dies for his name-stamp, and his mortaria exhibit a wide variety of rim-types and fabrics; there can be no doubt that his period of activity was a long one. His stamps occur at Lincoln (British Museum exampleso ,tw ; cf. GIL, vii. 1334.9), Aldborough (Reliquice Isuriance, pi. xxxiii.), Corbridge (many examples, as yet unpublished) and Lanchester (unpublished) in the north of England; and at Newstead . 266(p , 28), Rough Castle 52). Hillr . 70)(p Ba ,(p , Balmuildy (pi . B.xl , 11), Camelo d Ardocan n h (NMA, unnumbered n Scotlandi ) . Haverfield EXCAVATIONS AT BIRRENS, 1936-1937. 331

refers, with reserve, to the reported discovery of a stamp of this potte kila t Hartshiln na i r Warwickshirn i l e (VCH Warwicks, . 246)p ; the variety in fabric, already referred to, seems suitable for a potter working in forest country and moving from place to place in search of fuel. 4. FP 101; stamped GRATINi. I have not yet noted any rim on which the complete stamp of Gratinus occurs; in most cases the first part bees ha n ot i fimpresse rime th here,s ,n a do while once th r twic eo s i t ei letterfirso tw t s tha missinge tar contrasn I . Sarruso t , Grratinus exhibits little variety in fabric or rim-section. His stamps have been recorded at Templebrough (No. 5), Wilderspool (Warrington s Roman Remains, p. 64, misread), York (OIL, vii. 1331.52, wrongly inserted amon e amphorgth a stamps Corbridgd an ) e (unpublished Englandn i ) Scotlann i ; t New-da ; NMA11 d stead, an . 266 FR, (p 10 A ,, 2 fig1454 , .35 , 1461, 1462d an ) Balmuildy (pi. xl. 6).

I 2 Fig. Stampe34 . d mortaria. (\.) I have been unable to find the incomplete stamp ]IAR, referred to in PSAS, xxx. p. 186. ,3 It is desirable to add a note on the general question of mortarium- stamps, which has hitherto been unduly neglected.1 There have been two main tendencies: either to regard such stamps as having mainly local intereste productth s a f ,potter o s s merely serving thein ow r restricted markets, tako t r eo 2 case identitf so namf yo e between makers of mortaria and makers of Samian ware as evidence for the manufacture of mortari latter.e th y ab 3 Neither vie whollws i y incorrect cleas i t i r; that many potters did serve a restricted market, for example Satu(rninus) of Corbridge, whose stamp occurs there and at Newstead,4 but has not t beeye n noted elsewhere e potteth r o r: whose stamp IMS r 03^o - - or bot conjunction hi t presena s i n t restricte sitex si Scotland.n i so dt 5 Again graffite th , i fro Gfraufesenqua mL e reveal that mortaria were made 1 The best discussion which I have come across is that by Mr S. N. Miller, Balmuildy, p. 79. 2 So Haverfield in AA3, viii. p. 194; May, Warrington's Roman Remains, p. 60. Ritterling, Hofheim, pp. 310-11; Oswald and Pryce, Terra Sigillaia, p. 211; Macdonald and

Curie3 , PSAS, Ixiii . 527p . .

. AACf , viii . 194p . d Newstead,,an ; th 25 fige , latte.35 r stamp (NMA r 1477A fa a FR , s )i 3

complete4 r impression fro Corbridge mth thae publishee edi nth d drawing suggests. 5 Newstead (four), Rough Castle (one), Ardoch (two), Camelon (three); Balmuildy (two)d Ol , Kilpatrick (one) I hav. e las th examinet t threbu l e al dstamps ; drawing fuld an ls references wil givee subsequenb a l n ni t paper. 332 PROCEEDINGS OF THE SOCIETY, MAY 9, 1938.

many b y South Gaulish potters perioa n i , d when mortari Samian ai n t with. t therme Bu e vert case1w ar e f no apparenfe ywaro s e ar e t identity which will bear the test of a close scrutiny; and while a study e distributiooth f f mortarium-stampo n s e caswilth f o eln i ofte s (a n Sarrus, above) show a far more than purely local market, it demonstrates tha e stampth t s from second-century deposit e nortth f Britain ho i s n e productmusth e b t f Britisso h potters, even though there "were Samian potters of the same names working in the same period. And in many cases ther demonstrabla s ei e disparit datn yi e betweepottero tw e snth whose identification has been attempted; thus, the Sarrus who made Samian ware is dated to the time of Nero, while the mortarium-maker

belong Antonine th o st e period lattee th ; r probably worke Warwickn di 2 - shire, whil namesaks ehi beed eha npottea Graufesenquea L f o r . Even where ther s identiti e f periodyo ,y mean identificatioan y sb t no s i n certain; for example, the Flavian mortarium-maker (L. Atisius) Secundus, to judge by the distribution of his products and by his nomen,3 worked n Gallii a Narbonensis d cannoan , e identifieb t d wit e Secundushth f o Graufesenqua L e familia studento t r f Samiaso Albinus,n n ca ware r no ; Marinus and Matugenus, who made mortaria in the Flavian period, be identified with the contemporary Samian potters of La Graufesenque, for their wares often bear the additional stamp Lugduni factum— "mad t Lugdunum.ea effectn I " permissiblt , thenno s i t ,i generaliseo et ; each cas f appareno e t identity mus e examineb t s meritsit t n o dbu ; "where I have made such an examination, the case for identification has usuall ye ground s th falleclearli o t t I ny . desirable thae th t mortarium-stamps from Britain shoul e properlb d y collected thao s , t further light can be thrown on the other question, the distribution areas and working-places of the different potters; I have already obtained some very interesting results fro mpreliminara y stud thif yo s kindd an , e Scottispapea th y n hopla o r ho e t material before this Societe th n yi near future. Fig. 35. 1937; in and close to the water-channel, west of the tank (fig. 13 and p. 307 above). Much of a very large storage jar in hard, •whitish buff ware s heighit ;s 17^wa t inches maximud an , m diameter 12^ inches. Below the rim is a frilled strip; there is a double cordon round the neck, and a single one just above the lower attachments of the two handles; the cordons have been roughly decorated by notches e shoulde t th wit cu n ho wheela paralleo d r thertw an ;e elar grooves,

lightly incise terminatind dan gshora t distance fro handlese mth e Th . 1 Cf. Hermet, Les graffiies de la Graufesenque, 1923, pp. 149-50 et al. Oswald, Index of Potters' Stamps, p. 281. Cf. CIL, xii. 764, from Aries. 2 3 EXCAVATION BIRRENST A S , 1936-1937. 333

base is well formed, and has a groove demarcating the footstand on the under side I hav. e foun o vern d y close parallel o thit s s handsome vessel t ther bu e som, ar e e fragments wit ha similarl y notched double cordon in the Corbridge collection, and cf. Niederbieber, type 73A, fig. 42

E/B. Fig. (i.35 . ) (a rather squatter vessel, with three handles, and somewhat similar incisee shoulder)th dn o d cordon Caerleon,d an , an sm ri belo e wth Archceologia Cambrensis, 1932, fig. 57, 118 (only about two-thirds the e Birrenth siz f o e s jar, rather differen formn fabri i n t i wit t d bu , hcan a frilled strip belo e rim) we fabrith th ; c seem t unliksno ee thath f o t Colchester face-urns, which normally have similar frilling s findIt . - spot shows that this jar can be assigned to the last structural period at Birrens. Early fourth century. 334 PROCEEDINGS OF THE SOCIETY, MAY 9, 1938.

Fig. Graffit36 . potter'd an i s stamp Samian so n "ware. 1. 1936; Site VIII, Level III+ . Owner's name, GAI—"the property of Grams," scratche e Samia e walth th f o ln do n cup. ,2 fig , .28 2. 1937; Site XXII West, overlying the filled-in ditch and sealed by later walls (cf. p. 300 and fig. 30, where other pottery from the same deposi illustrated)s i t . Owner's name, apparently [VJITALIS, scratched on the under side of a base fragment from a Samian platter of form 1.8/31. 3. 1936; Site VIII, Level III + . 'On a platter of form 18/31, burnt blaca woo n ki d e letterfireo firs a th potter'tw t:f o s s stamp, AE[. There are too many second-century potters whose names begin thus,

I 234- Mg. 36. Graffiti and potters' stamps on Samian ware, (f.) foattempten a r d restoratio name profitablee b th f o neo t . PSAScf ; , Ixv. pp. 433-34. e stamth forf , o (fig p1) 3 m p 3 , cu .30 a 4 . n 1937O . 2 ; . origiNo s na GNATI • M. Gnatius is assigned conjecturally to La Madeleine by r Oswal D Indexs hi n di of Potters' Stamps, . 138p ; three other examples stamps ohi f , non thef eo m complete, have been recorded from Scotland (PSAS, Ixv. p. 437); the forms made by him justify Dr Curie's attribu- Antonine wors tioth hi f o knt o e period (Newstead, wit9 . 236)h22 p . t p ,bu there do not seem to be any records of his stamp occurring in stratified deposits. C. Glass. Only one vessel was represented by a fragment large enough to permit a drawing being made; that was part of a hemispherical cup, in colourless cut-glass (fig. 37), foun 193n di Sitn 6o e VIII, Levefacete I Th +n lII so . e rathear thip r cu smor e squa e tstraight-side th tha n no d beakern i , similar technique, foun t Birrena d 189n i s 5 (PSAS, xxx . 109)p . . Cut- glass vessel thin si s styl decoratiof eo n hav ewida e distributiono d t bu , not seem to occur very commonly anywhere; cf. Dr James Curie's valuable discussion in PSAS, xxx. pp. 110-1, where a date late in the Roman period is suggested for the Birrens beaker. Since 1895, evidence has come to light which suggests the possibility of a considerably earlier EXCAVATIONS AT BIRRENS, 1936-1937. 335

date; thus, Oelmann, in his monograph on the pottery from the Roman fort at Niederbieber in Upper Germany,1 discussing this type of decora- tion, quotes a number of glass vessels of closely related form and assigns them to the latter part of the first century ; and an extremely close

parallel to the 1895 piece has been found in a2 sealed deposit antedating the Trajanic reconstruction of the legionary fortress at Caerleon.3 But the evidence cited by Dr Curie is sufficient to demonstrate the persistence of the type into the late Roman period, and the stratification of the present piece doe t accorsno d wit earln h a bee d y ha nstradate a t i f yi : from an earlier level, it could hardly have escaped breaking into fragments, for it is thin, delicate ware; it is probable, therefore, that it belongs to

~)

Fig. (J.37 . ) vanishee th d Leve , likmortariume IV eth l e cookingth d an ,- 1 fig , .33 pot, fig. 28, 7, and should be assigned to the third century. The glass from the burnt deposit on Site VIII, Level II, included part e footstanoth f a platte f o d colourlesn i r s glass d veran ,y many pieces from one or more rectangular bottles in greenish glass. It has not yet been possibl t togethepu o et r enoug fragmente th f ho shoo t s fore wth m of vessel represented, but there is a handle similar to that of Hofheim, type 12,4 and there are at least three plane surfaces ornamented with two concentric circles, flanke singly db e upright trees.

D. Bronze Objects. hopeabld e b includo ha t e o t dI a eful l discussio e bronzth f o ne r Felloobjectou H . r Kilbride-Jonesy wE .M b s , but bees , sincha ne eh prevente pressury db f otheeo r work from completin s contributionghi , it has become necessary for me to provide a brief description, which I hope that he will be able to supplement 011 another occasion, in the 1 Niederbieber, . 8;p cf. also PSAS, lii . 219p . . e evidencTh e cite y Oelmanb d n doe t appeano s r decisivee Caerleoth t bu , n parallel proves 2 his" point. 3 ArchcBologia Cambrensis, 1929, fig. 18, 2. 4 Hofheim, pi. xxxviii. and p. 373; the Hofheim type, however, is not rectangular but cylindrical. 336 PROCEEDINGS OF THE SOCIETY, MAY 9, 1938.

J.A.VTANFI£U) Fig. 38. Bronze objects. (\.) EXCAVATIONS AT BIRRENS, 1936-1937. 337

courswidee th f eo r studbronzee th f yo nativf so e manufacture whics hi so badly needed. Wit exceptio e unstratifien hth a , 4 , figf 38 no . d find mad Sitn eo e XXII Easfollowine 1937n th i t l al , g objects were e burnfounth n i dt deposit, SitLeven o e e , VIIthereforII e lth ar n 1936 n i Ii foun e d b w an ,o edt no Dumfries Museum. Fig. 38, 1 and 2. Terret rings, ornamented each with three square bosses once inlaid with enamel, traces of which remain, let into the geometric framework illustrated in fig. 38, 1 A. A third ring of the same type and probably from the same set was also found, in fragments. The type is a well-known example of native metal-work; cf. Newstead, p. 302; the decoration invites comparison with that on other categories of native craftsmanship. Fig. 38, 3. Cheek-piece for a bridle; on the side is a decorative feature in relief to which, as Mr Kilbride-Jones points out to me, there are many counterparts on a variety of bronze objects found in the north of Britain. Fig. 38, 4. Indeterminate object, perhaps a weight. . 1 Ornamenta , Fig39 . l stu attachmenr dfo harnesso t t ; thers eha bee na centra l decoration missing w e circulano th , s a , r depression into which it fitted shows. . 2 Handl , Fig39 . e fro ma smal l bronze jug morA . e elaborate counterpar o thit t s handle, fro e Ruberslamth w hoard foun n 1863i d , is illustrated in Dr James Curie's paper on objects of Roman and Roman provincial origin, PSAS, Ixvi., 1932, p. 367, fig. 51. Unlike the preceding items, this object is not of native manufacture; parallels may be found throughou Romae th t n world. Fig. 39, 3. Bronze shoe for a wooden pole or shaft. It is hardly necessary to stress the interest of this group of objects from Site VIIIhomogeneoua s a , s deposia f o t d assignablen e th o et period whicshows i s h(a n elsewher thin ei s paper) closed shortly before A.D. 158; but it seems worth noting that the terret rings and bridle- piece, while they attest the presence of mounted men, need not be taken to show the presence of a cavalry regiment at Birrens: for both the second cohort of Tungrians and coh. I Nervana Germanorum (the only other regiment known to have been in garrison there at some time 1) were equitatce. 1 OIL, vii. 1063 and 1066. 338 PROCEEDINGS OF THE SOCIETY, MAY 9, 1938.

J.ASTANFItU)

. FigBronz39 . e objects. ({.) EXCAVATION T BIRRENSA S , 1936-1937. 339

E. Inscription. The only inscribe e upped th ston s r ewa parsmala f o t l altar, found in 1937 partly protruding abov e turSitn eth o fe XXII West. The letters (fig. 40) are not deeply texe th t d presentcutan , difficulta s e lasth t n i y symbol in line 2; but it seems possible to obtain NEP7 followinthe g reading: d(eo) Nept\un\o Cl(audius) [• • • —"Claudius . . . (fulfils a vow) to the god Neptune. t I nee" d occasio o surprisn o fint e d Mg. 40. a dedication to Neptune at an inland site like Birrens; compare Sir George Macdonald's observations on the altar, similarly dedicated, from Castlecary.1

F. Coins. Only one coin was found in 1936, a bronze one so far decayed that nothin e excavationg Th coul . e madit b d f f 193o o es 7 were more fortunate, producing eleven coins in all (as many as are recorded in the repore excavationth n o tthef o l m f 1895)al so wer t poobu n ei ; r condi- tion onld coulx an ,y si deciphered e db . Perc indebteW m a r yI M . o dt Hedlefollowine th r yfo g list:— No. Find-spot. Denomination. Identification. Date. 1 XXII East. Denarius. , 98 Traja . (M n A.D. 103-111. C. 120). 2 N. rampart. Denarius. Trajan (M. 331, A.D. 114-117. C, 190). 3 XXII East. Sestertius. Trajan. 4 XXII East. As. Trajan. 5 XXII Westn i , As. Antoninus Pius A.D. 154-155. black layer. (M. 934, C. 117). 6 XXII East. As. Antoninus Pius A.D. 154-155. (M. 934 . 117)C , . 7 N. rampart, below As. Illegible. outer cheek. 8 XXII West. As. Illegible. 9 XXII West, in Denarius. Illegible. black layer. 10 XXII East. Denarius. Illegible. . . oXV S f 1 1 Denarius. Illegible. It will be seen that no further light is thrown on the occupation of thesf o y e an coins fore f the i y beed th , b t y9 ha ; n d Nosdecipheran 7 . - 1 BWS . 4222 p ,. als cf ;o Domaszewski, Abhandlungen romischenr zu Religion, 1909. 19-21pp , . VOL. LXXII. 23 340 PROCEEDINGS OF THE SOCIETY, MAY 9, 1938.

able, might have been of value, but the coin of Antoninus Pius, coming from the same layer as No. 9, is clearly a good deal earlier than the termina le deposi datth f e o founs whic n i t wa dt hi (cf.. 346p , below). t remainI mentioo . sCunninghat A r coino M n tw y s b show s mu o nt of Larchcroft, Ecclefechan; they are said to have been found during the excavation workmee th t f Birrensa o e n 189n on si employe y 5b d there, from whose daughter Mr Cunningham acquired them. The first, a worn bronz Domitianf eo e ordinary th s nothinsecone i , f th n o a t t s di bu g, ou antoninianus e Gallith f o c usurper Victorinus, A.D. 268—70. Since th e pottery foun 193n 193i d 6an 7 includes material sufficien shoo t t w that s occupiee forth wa t d throughou e thirth t d century, neeo thern ds i e to question the attribution of this coin to Birrens, though it must be admitted tha credentials it t s might have been better. Mr Cunningham also showed us a coin found about 1935 in repairing a stone dike at Dockenflat, near. Ecclefechan, which it will be convenient placo t recorn eo bronza d s i here t I e .coi Maxentiusf no minn i , t condi- tion; obverse, MAXENTIVS P F AVG, reverse CONSERV VRB SVAE, mint-mark PT. VII. CONCLUSIONS. A correlation between the structural periods observed at Birrens and the successive phases of the Roman occupation of the north of attainable b t ye t Britaieno wity hnma certainty t seemi t bu ,s necessary attempo t t one f onli , shoo yt whawn i t respect r evidencou s e requires supplementede b o t bees ha n t I show. n that ther stils ei l uncertainty, at Birrens itself, as to the precise relationship between the periods of the rampar e internathosd th an tf eo l buildings betweed an , e levelnth n i s unenlargee th d northware thosford th an tn ei d extension followine th ; g attempt to define the relationship, and to connect the various periods with the general history of the Roman occupation, is put forward with l reserveal . e firsth t n placeI t wil i ,convenien e b l emphasiso t t distinctioe eth n between the evidence relating to the north of Britain in general, and that derived from Birrens itself, for the historical framework into which the structural periods have to be fitted; the latter category is the simpler, and may be summarised first. Apart from a piece or two to suggest the possibilit Agricolan a f yo n occupation, e potterth 1 y series starts with time th Hadriaf eo thire continueclose d th th nan f d eo o centuryt p su , while one late third-century mortarium,2 underlying the water-channel insid northe eth rampart, shows tha latese th t t structurale periob y dma 1 Cf. pp. 290, 321, above. 2 Pig. 31, 14. EXCAVATION BIRRENST A S 1 , 34 1936-1937.

assigne s lat a s circa ea o dt A.i>. 300. Withi e perionth f nearldo o ytw hundred years so defined, there are as many as four periods of stone t buildingdoe i t belonf e Agricola(i no sth d o gt an s n occupationf o e on ) wood, to be correlated with historical changes; but while the stratified pottery, as has been shown, suggests a partial time-table, Birrens itself has only produced direct evidence for the date of one such change, namely, the inscription of A.D. 158. In order to resolve the problem, • we must turn to the general evidence already referred to. The complexio e potterth f no y series allow begio t r surves u snou y wit e buildinhth f Hadrian'o g s Wall, circa A.D. 122 t thaA . t tima e number of forts were constructed, not in immediate touch with the new limes: particularn i e outpostth , t Bewcastlea s e , w Netherb s (a d an y positioa n i w addaro nt eno ) Birren notee b y ds thaitselfma e t I th t . excavations recently conducte t Bewcastlda Richmonr M y eb d shoe wth Hadrianic fort there to have had buildings of timber,1 like the mile- castles westward from the Irthing on Hadrian's Wall,2 and perhaps the first fort at Castlesteads; 3 this might be taken to support the view that e initiath l timber perio t Birrenda s shoul assignee b de tim th f o eo t d Hadrian and no earlier. The next stage comes wit e reoccupatiohth norte th f Britainf hno o , beyond an uisthmue o pt d th s between Fort governoClydee d th han y b , r Lollius Urbicus in the early years of Antoninus Pius; at this time stone forts, later retained as outposts of Hadrian's Wall, were built at Rising- ham and High Rochester: there is considerable evidence to show that a

rearrangemen f garrisono t s too e Hadrianik 4 th e fort placth f o sn i e c frontie conceivabls ri t zonei d :an 5 e that sucn o ,occasionn ha y ma t i , have been found convenien replaco t t e timber building buildingy sb f o s stone in a fort like Birrens, to suit the requirements of a fresh regiment. But it should be noted that no such change occurred at the nearby fort of Bewcastle. After the governorship of Lollius Urbicus, the available evidence falls intgroupso otw , associated wit e Hadrianihth d Antoninan c e frontiers respectively. Hadrian's Wall.6—Here e firsth t period, inaugurated twenty years previously, continues as far as the forts are concerned without interruption unti secone close th lth f eo d century, whe ndestructioa n occurs whicn hca now be assigned to the year 196, when Clodius Albinus was defeated in Gau y Septimiub l s e latter'Severuth d san s first governo f Britaino r , 1 CW2, xxxviii . 195-237pp . . 2 Cf. JUS, xxv . 1-18pp . , especiall. 8 . yp 3 CW2, xxxiv . 164p . . * AA4, xiii . 170-98pp . . . JRSCf « , xxii . 55-59pp . ; AA4, xii . 199-200.pp . . CWCf « 2, xxx . 199-202pp . ; AA4, vii. 167-69pp . . 342 PROCEEDINGS OF THE SOCIETY, MAY 9, 1938.

Virius Lupus, found it necessary to buy off the Maeatse, who had created widespreae th d havo whico ct h each fresh excavation bears added testi- mony secone Th .d period opens wit reconstructioe hth Viriuy nb s Lupus and his successor Alfenus Senecio, and closes in a comparable destruction at the end of the third century, when the usurper Allectus was engaged in his unsuccessful struggle against the Csesar Constantius Chlorus; and the building-record found at Birdoswald in 1929 allows us to place the beginning of the third period circa A.D. 300, as the work of Constantius Chlorus himself. Thus far Risingham, High Rochester, and Bewcastle can be shown to have experience same dth e historical sequenc Hadrian's ea s n Walli t bu , the fourth century a divergence occurs. The third Wall period ends in destruction by the Picts in A.D. 367; at Risingham and Bewcastle the pottery series reache (buo t t beyondp tsu no ) that year t thern a bu , s ei intervening destruction followed by rebuilding, which it seems justifiable assigo expeditioe t th o nt f Constanno A.Dn si . 343 t Higa ; h Rochester, on the other hand, the pottery series does not extend so far, there is no such rebuilding, and it seems that the site was omitted from the recon- struction carried out by Constans. In view of the lack of late pottery from Birrens, it may well be that it shared the fate of High Rochester; we wil justifiee lb puttinn di g forward circalatese th s ta A.D dat0 r .34 efo its occupation by the Romans. But whereas High Rochester has three structural period coveo st yeare th r s from circa 140-340, Birrent a s ha s least four and, as we have seen, may have five. If it had been four only, it would have been a simple matter to suggest that the rebuilding in A.D. 158, without a counterpart at High Rochester, explains the difference; but we must retiirn to the question presently. The fourth and closing period of Hadrian's Wall, opening with reconstruction by Count Theodosius. in A.D. 368-9, and continuing at least unti e usurpatioth l 'Magnuf no s Maximus knowo n s nha , counter- part on any site to the north of the Wall, so that it need not detain us further. The Wall f Pius.o —When e tur w o conside nt e histore th r th f yo Antonine frontier, we are on sure ground in stating that it falls into three structural periods, but the attribxition of those periods is open to question. mose Th t recent vie ws tha i t forwar r Georgpu t Si y b de Macdonaldn i , the second edition of The Roman Wall in Scotland,^ where the periods are assigne e followinth o dt g time-table , circaI : A.D. 142-155/8 , circaII ; A.D. 158-181; III, circa A.D. 184-185. But that view is so largely based 2 on an interpretation of the history of Birrens which has been shown by 1 BWS . 478-822pp , . 2 BWS . 4782p , . EXCAVATION BIRRENST A S 3 , 34 1936-1937.

the recent excavations to be untenable, that it will be necessary to consider the evidence afresh. We may start our reconsideration with a statement of principles. In any case of conflict between literary and archaeological evidence, the latter clearly e examineneedb o t s d with great care b e n beforca t i e allowe overrido dt e conflicth e formerf i betwees e th i td an ; n positive literary evidenc negative th d ean e evidenc archaeologyf eo t wili , l neen da overwhelming argument to justify disregard of the literary record. Such a conflict exists in relation to the Roman occupation of Scotland, and the negative archaeological evidence has been generally permitted to override that of our literary sources. Briefly, the key-stone of the current archaeological interpretation is e absencth f coineo s later tha e tim nth f Commodu eo s fro l Romamal n fort Scotlandn si , wit exceptioe hth Cramondf no ; this absenc hels ei o dt justify the view that, with that exception, none of those forts was occupied in a later period. At first sight the argument may seem a sound one;

t shouli t rememberee bu db d tha e totath t 1 l numbe f coino r s froe mth Antonin usefue b e cas t vere poiny Walo t lth no e yma s o li t greatt i d an , of Housesteads fort on Hadrian's Wall, where the excavations of 1898 produce coins 9 mans da 12 s ,t yinclud a whic no d h di esingla betweee eon n time th Commoduf eo thad Elagabalus;f san o t t thaye t fort continued

in Roman hands until2 the close of the fourth century, and it has produced fragments of a Severan building-inscription.3 And when we turn to the literary evidence, which has been unaccountably neglected, it seems to mak t ecertaii n that Roman Scotlan reoccupies dwa t leasa r tdfo four years, from A.D. 207 until 211. e formeth n I r year s Cassiua , records,o Di s 4 Severu n Romi ss wa e greatly disturbed because he was unable to put a stop to the activities of a notorious brigand in Italy, at a time when his generals were winning victorie Britainn si scene Th thosf eo . e victorie t recordedno t s i si t bu , n hardlca y have been sout f Chevioto e hbuilding-recor th r fo ; d from Risingham, dated circa A.D. 205,5 shows that reconstructio alreads nwa yn i progress there—and it is perhaps significant that the work is described as being unde superintendence th r procuratore th f eo , Oclatinius Adventus. e inferencTh e seems justified thae governorth t , Alfenus Seneciod ha , entrusted the 'work of reconstruction to the procurator, while he himself followe e tid f th dcampaignineo g further north. Therquestioo n s ei n that it was farther north that Severus and Caracalla conducted their campaigns agains e Caledonianse Maeatath th t d an e , wit e avowehth d

1 BWS2, p. 480; PSAS, lii. p. 275. "- AA2, xxv. p. 298. 3 AA4, ix. pp. 233-34. « Book Ixxvi. 10. 5 OIL, vii. 1003. 344 PROCEEDING E SOCIETY , 19389 TH Y F O .S MA ,

object of conquering the whole of Britain ;J and, after the death of Severus in A.D. 211, we learn that Caracalla made peace with the enemy and with- dre garrisons'froe wth m their territory.2 . When we find the archaeological evidence from the forts on the Antonine Wall showin a gbrie f final occupation, endin n peacefui g l withdrawal t seemi , s difficul resiso t t conclusioe th t n that this occupation shoul assignee db e withdrawao Severudt th d an s o Caracallat l e th s a , literary sources suggest, and that the absence of coins of Severus is as muc accidenn ha t thert Housesteadsa s ea . n thaI e leftar t cas e wito structuraw e htw l period o covet s e th r interval betwee e Severath d n an A.Dn 2 reoccupation.14 ; and n vie,i w of the evidence available, it seems possible to suggest two alternative time-tables f UlpiuI ) (1 s. Marcellus reoccupie e northerth d n limesn i A.D. 184,3 the second period will presumably have ended in the same destructio s overtooa n k Hadrian's Wal A.Dn i l . 196d thaan ;t leaves A.D. 142—18e firsth tr periodfo 1 , which terminate e barbariath n i s n incursion recorded by Cassius Dio.4 In that case the reconstruction t Birrena A.Dn wil8 si counterparo 15 l. n hav d eha eithen o t r Walld an , the historical explanation for it must be sought in the immediate neigh- bourhood—a possibility t whicweakene no e geographs th i h y b d f o y the area. (2) If Marcellus contented himself with drastic punitive operations, reverted e Hadrianian 5 th o dt c frontier system, leavine gth Antonine limes unoccupied t liberta e n occasio a ar o loo yt r e fo kw , n beginnine th firse d th an tf o g d en betweee th r fo circand 1 A.Dan 18 2 .14 of the second period; and that occasion might well be contemporary wit e nee reconstructior hth dfo t Birrensna doubtee b , thougy ma d t hi whether there was any necessary connection between events in what Dumfriesshirw inorthere sno th n o d n limes.ean The Periods at Birrens.—It appears, then, that there were two structural periods in the forts of the Antonine Wall between A.D. 142 and the end of the century, as against the single period on Hadrian's Wall. High Rochester, Risingham e samd Bewcastlth an ,en i e ar e case as the latter, while Birrens, with the reconstruction of A.D. 158, seems to compare more closely with the former. It is time to consider "whethe arrivn closea ca t e ea w rr correlatio periode th f n o t Birrenssa . (1) The First Period.—This is marked by the timber buildings of which traces were found below SiteWesX d VIII XI sfavoun an tI . r of an Agricolan date is the apparent difference in alinement noted on

1 Cassius Dio, Book Ixxvi. 13. 2 Cassius Dio, Book Ixxvii .. Herodian1cf ; . ,6 iii, 15 . 3 Cf. BWS2, p. 479. 4 Book Ixxii. 8. 5 Ibidem. EXCAVATIONS AT BIRRENS, 1936-1937. 345

e latteth r sitee cooking-poe Dumfrieth ;th n i m s ri t Museum x seems clearly pre-Hadrianic e Samiath d nan , bowl from this level2 might belong to the same period, of which Curie 11 is one of the most char- acteristi othece typesth n r beeO s hand. ha n t i ,pointe t thadou t there neeo i sn postulat o dt edata e earlier than Hadria thar nfo t particular piece, and the fact that the Hadrianic forts at Bewcastle and perhaps Castlesteads wer timbef eo seey mjustifo ma rt y assignin firse gth t period to that emperor. Further deep digging, within the area where timber buildings have been noted, is plainly required to settle the point. (2) The Second Period.—The timber buildings are replaced by stone barracks in alinement with the existing ramparts; if the inference drawn fro positio mthe post-hole the nof Siton s e VIII correctis firsthe , t

period in the section3 through the west rampart, and the early north rampart, will belong to this period. The small yield of pottery from Level II on Site XIX West, and from the lowest level on Site XXII West, is consistent with an attribution of this period to the years circa 122—158, but here too further pottery is required to confirm the dating. (3) The Third Period.—This brings the reconstruction of Level III on Sites XIX West and VIII, the latter yielding a fair amount of late second-century pottery; 4 the material found underlying the visible north rampart 5 justifies placing the extension of the fort, and with it the second rampart-period in the west section, and the first stone buildings on Site XXII thin i , s period datee , b whic y d A.Dhma . 158-circa 196. t shoulI observee db d that pottery evidence fro firse mth t perioe th f do northward extension is still badly needed; at present the platter from Site XXII Easonle th ty s 6i piece . (4) The Fourth Period.—This period opens with the Severan recon- struction e deposith s a , t sealed below West, X Sitn Leveo XI e V I l

and pieces from Level III + o7 n Site VIII8 and the second level on Site XXII East,9 allow us to infer. It seems reasonable to assign the thicken- refacind in e nortgan th f hgo rampar same th eo t t period,10 particularly because there is a later structural phase there, which must be reserved for the final period. (5) The Fifth Period.—The beginning of this period is to be assigned to Constantius Chlorus, on the strength of the mortarium found under- lying the water-channel near the north rampart,11 and the cooking-pot 1 Cf 280. p . , above. 2. Fig1 , 26 . ' Cf . 292p . , above. , 2-3 d fig33 .an . 8 . fig2 .Cf 4. 29 d an 5, Figs23 . , « 13 Fig22 . , .31 7 'Fig . 27 . . 1 , 33 d 8an 9Figs7 . Fig, 9 , 28 . .31 . 305p ,. 1Cf above0 e Samiath ; n rim , whic10 , , figs clearlhi 31 . Antoninn ya e type founs wa , d in the bottom of the turf revetment at the outside, where it appears to represent a refacing; this waonle sth y stratified evidenc secone th r edfo perionorte th f hdo rampart remainine th r ,fo g pieces were sealed by its original structure. u Fig. 31, 14. 346 PROCEEDINGS E SOCIETY O, 1938TH F9 Y . MA ,

rim among the debris overlying Level IV on Site XIX West. Much

of the "secondary" work detected in 1895 probably belong1 s to the same period, whose maximum extent may be put as from circa A.D. 300 until 340e aqueduct e th thir; th dd e perionortan , th hn i d rampart, belonsame th eo g t stage. The situatio n Sito n e XXII West still presents some difficulty. There we have three surviving levels subsequent to the extension of thesf o e secon e fortd eth th s producean , dha e potterth d y illustrated as figs. 24, 3 and 4; 31, 1-8; 32, 2-4; and 33, 9. The detailed discussion of that material show s parallele e e piecetha b e bulth n th tf kca o s n di the latter part of the second century rather than in the third; yet above onth e consideratio periode th f n o t Birrensa firse sth t stone build- e underlyiningsth t a , g level, ough o belont t thao t g t period. There alternativo tw e ar e explanation bees ha t nforward t I notepu ) o st (a d, that no pottery was found at the same level as the first stone buildings,2 d thaan t they were largely reduce theio dt r clad cobblyan e footings. It seems possible that, at the Severan reconstruction, all the structures in this area were demolishe debrie theth s d snwa d an sprea d evenly over the site before fresh buildings were put up, the top layer of the spread being composed of occupation material from a neighbouring building. (6) Less likely is the suggestion that the pottery should be taken to show the persistence of the types concerned into the third century, though there are certainly one or two pieces for which a third-century date seems preferable.3 Here, too, further diggin s urgentlgi y needed to settle the problem. Summary.—The occupatio d bee t ha f endBirren no nno s a , d di s supposed, befor e secone closth th ef o et di confine s centurywa r d no , structurao tw o t l e periodscontraryth n O . . , there were five such periods, carrying its occupation at least from the time of Hadrian, and perhaps from the governorship of Agricola, until some time in the first half of the fourth century. Further excavation is required to fix the apportion- ment of those periods to' the phases of the Roman occupation, and to e structurath l sequenc e ramparn differenth o n d i e an tte th site n i s interior; but it seems justifiable to put forward the following provisional time-table:—

Period Ifora : t with timber buildings differena n o , t alinemeno t t visiblefore w th no t . Circa. A.D80 . Period firse IIth :t stone fort, with turf rampart lackin gfoundationa , over fifty feet shorter than its successor. Circa A.D. 122-158.

1 Pig. 26, 4. * Cf. p. 299, above. 3 Fig. 32, 2-4. EXCAVATIONS AT BIRRENS, 1936-1937. 347

Period III: the fort enlarged, and the buildings not within the extension reconstructed. A.D. 158—196. Period IV: reconstructio ramparf no internad an t l buildings. Circa A.D. 205-297 (=Perio Hadrian'n o I dI s Wall). Perio : furthedV r reconstruction e provisioth e waterd th an f , o n- supply system recently detected. Circa A.D. 300-340 (=the penultimate period at Bewcastle, the last period at High Rochester, and the first part of Period III on Hadrian's Wall). And, in view of the considerations advanced above, it seems likely that the explanation of the reconstruction required in A.D. 158 is to be sought in the history of the immediate neighbourhood rather than in the main course of events affecting the Antonine Wall, where two structural periods of second-century date are to be observed, or Hadrian's Wall, where there is only one.