Proc SocAntiq Scot, 124 (1994), 299-314

The provision of fort-annexes on the Geoff B Bailey*

ABSTRACT This paper suggests that temporary bathhouses were provided Antoninethe in Wall forts until such time as a planned could be completed behind the Wall. The protracted construction period caused many theseof bathhouses modifiedbe to provideto utilitiesthe required during this prolonged interval. The Wall had been built to before the concept of a Vallum was abandoned and consequently annexes were attached to each fort. At this stage it was then possible to demolish the internal bathhouses and to build larger bathing complexes in these enclosures.

INTRODUCTION This papeattempn a s i r provido t explanatioea severaf no l perplexing phenomena connected with Antonine th e Wall: Antonine Wall fort extremele sar y unusua westere th n li n Empir possessinn ei g internal bathhouses (Johnson 1983, 194); they have annexes rather than a Vallum like that of 's Wall (Salway 1965, 158); most of them apparently had their defences 'slighted' c AD 155 (Steer 1964); and they have different ways of attaching the annexes according to their location along the Wall (Breeze 1984, 61).

THE BUILDING SEQUENCE ON THE WALL Antonine forte th f Th o s e Wall vary considerabl individuan ow s it sizeacn d yi s elan h ha interna l layout. Yet, ther significana s i e t degre f consistenco e y in their planning, enoug indicato ht a e unified scheme. Of the 16 known forts all appear to face north with the exception of (which faces east owing to the local topography) and the two terminal forts which both face away from the Wall. The principal buildings of these forts - the principia, granaries and commanding centrae officer'th n i l y rangs la housl eal (excepe- t Bearsdena t mosn i d t an case )ston d ha s e foundations. The location of the granaries within this central range is particularly regular, with one on each side of the principia when there were two, or at one side of it when only a single granary was provided (see note 1). This situation is typical of the Antonine occupation of and may alsforte seee th oNewsteaf b s o n n i Higd dan h Rochester. Thisharn i s si p contras forte th so t of the Flavian period when the granaries were usually paired together and placed near to one of the principal gateways (eg Elginhaugh, Fendoch, Pen Llystyn). In the interval between these periods their locations had been quite varied (Gentry 1976). It is this variety which is displayed on Hadrian's Wall but is lacking on that of (cf Breeze & Dobson 1970).

' 412 Bank Street, Slamannan, FK1 3EZ 300 SOCIETY OF ANTIQUARIES OF SCOTLAND, 1994

This Antonine arrangement of granaries within the fort is also known in Germany (Gentry 1976). Wha s evei t n more unusual abou Antonine th t e Wal le coheren fortth s i s t provisiof no internal bathhouses (illus 1; Johnson 1983, 194). In most cases these lay between the intervallum road and the rampart, the result probably of shortage of space, although at Cadder, where space seem havo st e been plentiful bathhouse th , e occupie angle dth e betwee norte easnd th han t rampart streets e bathhouseTh . s were built afte e rampartsth r evern i t ybu , case where they have been examined the believee yar contemporare b o dt y wit othee hth r fort buildings. Elsewher Roman ei n Britain internal bathhouse e onlar s y foun secondarn i d y contexts, normall lata f eyo periode Th . elaborate bathhouse complex in the fort of Haltonchesters, for example, was a fourth-century imposition (Daniels 1978, 86), and similarly, the internal bathhouses at the Welsh forts of Brecon Gaer and Caernarfon were not original features (Jarrett 1969, 51, 169). An unusual local example is provided by Strageath where a second-century bathhouse occupied a position next to the rampart simila Antonine thoso th t r n eo e Wall. Here buildine th , s clearlgwa insertion ya a t liei t s a na strange rampare anglth o streetd et an t , partly blockin e lattegth r (Frer Wilkee& s 1989, 98). Normally bathhouses were built outsid fortse eth , annexoften a Gelligaet n a ni s ea r (Ward 1909). The known internal bathhouses on the Antonine Wall all began life as very simple structures, essentially consisting of three rooms in a row, representing the changing room (apodyterium), the warm room (tepidarium) and the hot room (caldarium). These bathhouse blocks were hypocausa characterized an d othee en entrancn t a th flue t y ra don e b (illut a e. Thei 2) s r size, locatio simplicitd nan l suggesyal t that haveacy hma e been intende temporara s da y measure, lasting perhaps only a few years. This indicates that the decision to provide internal bathhouses was made at an early stage in the building programme of the Wall. On Hadrian's Wall the fort bathhouses lay outside the forts by convenient sources of water, often some distance away. They wer t enclose eno protective th y db e shielannexen a f do t the bu ,y gaid di n securit theiy yb r emplacement withi militare nth y zone bounde norte Wale th th h o t ly db Vallue e southth th d effectmo n t I an . Vallue th , m actegiana s da t annexe extending alone gth whole length of the Wall (Salway 1965, 158). As Hadrian's Wall most probably provided the Antonine modeth r intentione fo lth s originae th Walwa musn e t i ,i whethek w l t las t no r o r scheme provido t , lattee eth r wit similaha r Vallum? Wer temporare eth y internal bathhousee b o st replace mory db e permanent external bathhouse latea t sa r date onc Valluea beed mha n dug? throwe Lighb n thesn ca nto e question bathhousee looke th on t sa f si s which replaced those insid fortse eth . These were much large mord an r e sophisticated bath suite wered san placee th n di fort-annexes. These annexe therefory sma e provid wits eu h another mean f testinso buildine gth g sequenc installatione th f eo s which for mAntonine parth f o t e Wall frontier, froe mb y whicma t hi possible to determine the place of the Vallum in the initial building programme. Were the annexes an afterthought? They almost all lie on the side of the fort which is away from, or upstream from, an adjacent burn or river. (A notable exception to this was , where the probable annexe lies on an area of ground prone to flooding and thus unsuitable for the fort itself.) Could these locations have been a consequence of the original position chosen for the isolated forte strongesth : t location topographically ,drinkins nexit o t t g water face ?Th t thae th t Antonine Wall curtain forms the northern rampart of the annexes shows that they have to be either contemporary with the Wall or later, whereas the northern ramparts of many of the forts can be shown to be earlier than the Wall curtain. (At Cadder the annexe is usually placed to the south of the fort, but this is not certain and the location of the external bathhouse makes it unlikely.) At , Rough Castle, Castlecar Duntoched yan annexe th r e ditches respec Wale th line f lth teo curtain, even though they sometime littla t esge close t Bearsde (illuA . 1) s eastere nth n ditches fall wel lWall e shorth f .o t Onl t Rougya h Castl t beei s neha possibl examino et junctioe eth n between BAILEY PROVISIOE TH : FORT-ANNEXEF NO ANTONINE TH N SO E WAL I 30L 1

150 ILLUS 1 Plans of the Antonine Wall forts; bathhouses are shown solid 1 Mumrills; 2 Rough Castle; 3 ; 4 ; 5 Croy Hill; 6 Bar Hill; 7 Cadder; 8 ; 9 ; 10 Bearsden; 1 1 ; north p to e th o t s i

stone th e foundatio Wale th f l no curtai annexe thath d f nan o t e rampart. Unfortunatel evidence yth e s ambiguousi . Althoug southere hth nWale kerth f l bo curtai n continued uninterrupted a 12-in, . (0.3 m) culvert placed alongside of it makes it impossible to determine whether or not the annexe rampart was contemporary with it, or later. Unfortunately the answer was not obtainable from the superstructure at this point (Christison, Buchanan & Anderson 1905). Earlier features, however, were found under the south rampart of the annexe (Macdonald 1933, 253). I SOCIET 2 30 ANTIQUARIEF YO SCOTLANDF SO , 1994

CADDER

L? f

BALMUILDY

MUMRILLS

BAR HILL

50

10 15 20

BEARSDEN ILLUS 2 Plans of the internal bathhouses showing their development BAILEY: THE PROVISION OF FORT-ANNEXES ON THE ANTONINE WALL 303

Fortunately relationshie th , clarifies pi examination a y db fortse th f 'no ditches. Startinn gi ease proceedind th an t g westwar looy t eackma a turne n hi dw t Mumrill A . annexe sth e ditch cuts the fort's three outer ditches (Steer 1961, 89). At Rough Castle the same sequence may be judged havo t e occurred detailo ditce n th t hf bu so , fill originae givee th ar s n ni l there reporty ar r ean no , sectional drawings at this point (Christison, Buchanan & Anderson 1905). At Balmuildy the annexe ditch stops well short of the fort's south ditches, indicating the latter's prior existence. At Wale situatioe th th f l o wese d th differents ni en t . Duntoche foro n ts ditcherha s withi annexee nth , instead they sweep aroun e soutdth hfore enveloo th t sidt f o e e annexepth . Her d fore eth an t annexe mus contemporarye b t . Bearsden, lying between Duntoche Balmuildyd an r , presentn a s even more peculiar situation where the annexe seems to have been carved out of the fort itself shortly afte buildine th r g programme ther beed eha n started (Breeze 1984, 61). An inspection of the ditch fills is illuminating and shows that the fort ditches were backfilled alon sidee gth s wher annexee eth s were added t MumrillA . three sth e outer ditche fore th t f werso e deliberately filled in, as is shown by the nature of the soil used and the statement that, of all the ditches onle th , y significant amount f pottero s y came from 'those which defende e westerth d n Antonine fronth f o tother e nond eth f foran eso - tyielde pottery dan speao yt f k(Macdonalo d& Curie 1929, 419, 502; Steer 1961, 91). These redundant ditches would have mad fittina e g receptacle for the fort's rubbish which would have included such pottery. As part of the construction work on the new annexe, gravel was extracted from the area around the ditches and waste from this proces uses swa d partlbackfilline th n yi g operation. Further materia obtaines wa l d probably by recutting the inner ditch on this side and the ditches on the other sides of the fort (Macdonald & Curie 1929, 417). Some of the gravel obtained from the area would have gone into the new roads laid out over the infilled ditches (ibid, 420) and throughout the annexe. These alterations thus brought the fort at Mumrills into line with other forts, with lesser defences on the side of the annexe (eg Pen Llystyn: Jarrett 1969, 102). The recent excavations at the fort at Falkirk have demonstrated a similar occurrence. At least two of the three eastern fort ditches which were available for examination had been deliberately infilled with clean sand containing some pottery fragments. This clean sand must have been derived from newl featureg vicinitye ydu th n i s , most probablannexw ne e eyth ditches. Ovee th r unconsolidate redundande th fill f o s t fort ditche timbee sth r building annexe roadd th s an f so e were lai dbuilkils a outd n wa superfluou t an w ,ove no ra s defensive slot (Bailey forthcomin. gb) At Castlecary it is also clear that the two outer ditches of the fort, on the side to which the annex s addedewa , were fille . Herin d e again excavatore th , s noted that only these outer ditches had produced pottery, and again a road was led across them (Christison & Buchanan 1903). By contrast innee th , r ditc founs h wa contai o dt n dressed stone possibly derived fro collapsee mth r do demolished fort wall thid an s, ditch would thus see havmo t e been retained. (The retentioe th f no inne infillin e t rRoug a outee th ditc e th d f hron an g o Castl plasuggestes S ei O n e produceth y db d befor 190e eth 3 excavations: Christison, Buchana Anderson& n 1905, 449.) Balmuildy has even more dramatic evidence, for not only were the two fort ditches (which were now in the annexe) filled in, but a bathhouse was built over them with a drain along the top (Millee oon f r 1922) late eannexeTh e . datth f hencd eo an bathhousee , eth , tie witn si theore hth y previously advocated, that the internal bathhouses were only temporary. As in all the other known instances annexe th , e bathhous larges betteed wa ran r constructed tha predecessors nit . That the internal bathhouse at Balmuildy did not remain in use as long as the fort was recognizee roomth f e excavator o sstreea th late e d y db on ha t r r laifo , d ove s demolisheit r d remains (Miller 1922, 46-7). The walls of the other rooms were levelled and rubble and clay were dumped int disusee oth d hypocaust system t Cadde suggestes A . wa t ri d thainternae th t l bathhouse 304 | SOCIETY OF ANTIQUARIES OF SCOTLAND, 1994

structure was adapted to change the function of the building; part of it was demolished and the remainder was altered to form a clubhouse (Clarke 1933, 59). There are also hints that the internal fort'e th f so lifd e en (Keppi e th bathhoust continuo t eno e r Hil1985d us Ba di l n t ei e a , 64)e Th . same pattern is repeated at Mumrills, where the internal bathhouse was not in use during the final phase of occupation, although the site of the external bathhouse is not known for certain (Macdonal Curid& e 1929, 462, 501) t CroA . y Hill contrasty b , externan a , l bathhous s beeeha n located (Macdonald 1932). Her dispute eth e mus ovee identificatiob te rth internae th f (see no on le note 2). Likewise, a building identified in the original report for Old Kilpatrick as a latrine (Miller 1928, 28) may actually have been a bathhouse (see note 3). Here the external bathhouse was cut through by the Forth & Clyde Canal (Miller 1928, 32). It lay in the annexe, near to the south-east corneforte fort'e th Th .f sro corner l roundedal e sar , indicatin buils g thawa tt i tbefor annexee eth . It is therefore extremely unlikely that the external bathhouse was an early feature, and may be seen as a replacement. The known duplication of bathing facilities seen at Cadder and Balmuildy (and possibly at Cro Kilpatrickd yOl vere Hild b yan y l tentativel)ma t forwarypu t Falkirkda . Excavation 199n si 3 located the main sewer outlet where it crossed the fort's ditch-system at its south-west corner. The fill of the middle ditch produced a fragment of a hypocaust tile which presumably was from a nearby bathhouse associated wit e latrinhth e block, theso nf I .thi s would make seconth t ei d bathhouse at Falkirk, for the external one was found some distance away at Kemper Avenue (Keppi Murrae& y 1981). Undoubtedly future research will brin ligho gt t more example f thio s s phenomenon. The situation at Castlecary is not properly understood as the internal bathhouse was not fully explored (Christison & Buchanan 1903), but in any case the known bathhouse was probably part commandine oth f g officer's house. Bearsden, once more, provides evidenc poine th t whicf a teo h all these radical changes appea originateo rt annexe th t onl s fore eNo .ywa th t carve f beforo t dou e s completefore th wa internate th t dbu l bathhouse, whic s thehwa n under construction behine dth rampart, now found itself inside the annexe (Breeze 1984b, 61). Instead of being finished as a simple internal bathhous expandes wa t ei d int largeoa mord an r e sophisticate annexe th f o ee don type wore Th . k thabeed ha tn doncompleabandones w ewa ne constructes a xd wa dan t righda t anglerampare th o st t (Breeze 1984b, 55). At Bearsden the life of this proto-internal bathhouse was extremely short, demonstrating that the builders were willing to abandon such small but well-constructed structures. However, most of othee th r examples display sign f extensioso improvemenr no t before the wero yto e abandonedf I . e postulatioth f planneno d obsolescenc s correctei t suggesti , s forte thar somth fo ta s f eo considerable time elapsed befor annexe eth added s same wa Th . e conclusio reaches ni notiny db g small alterations that were made to the fort ditches at Falkirk before they were filled in during the provisio annexee th outermosf e no Th . squars it d te ha ditc shaped hha d butt-end backfillea r dfo roundew ne distanca d m4 butt-enan f eo dugs dadditionn wa i ; north-ease th t a , t cornee th f o r middle ditch a cleaning slot at its base had been filled in and a new floor of clay was placed at a higher level (Bailey forthcomin t alon. Falkire e wholnorthno gb) th th this n s i er f eo -fo k wa , eastern ditch was redug at Balmuildy before the annexe was created (Miller 1922, 5). Dating evidence fro backfillee mth d ditche t Mumrillsa s suggest datsAa c Df eo 155-6 0 there (Steer 1961, 98). This would agree with the date of AD 155-60 given for a Samian stamp which provides a terminus post quern for the phase I occupation at Inveravon fort (Armit & Dunwell 1992) nature Th . f thieo s phase iron-smeltinn a , g establishment, indicates tha t representi t e sth fore th t initiatiof (contrao t annexe no th f d no Armi ean subsequen e Dunwell& th t o d r no ; t phases on the site give any reason to hint that this area did not remain part of the fort's annexe). BAILEY PROVISIOE TH : FORT-ANNEXEF NO ANTONINE TH N SO 5 E 30 WALL

These late dates, some 10 years and more after the work on the Wall had begun, reflect how lon t toogi evolvo kt finae eth lfrontiere forth f mo . Construction Wale worth ln ko curtain o d nan the forts had begun in the east, probably at (Bailey forthcoming a), and had gradually progressed westward. Shortly after work had begun at Bearsden it would seem that the plan to adopt annexes was initiated. Both the nature of the annexe and bathhouse at Bearsden point to this conclusion. Further west, at Duntocher, the annexe was contemporary with the fort which replaced a freestanding fortlet. Only once the Wall had been completed through to Old Kilpatrick would it have been necessary to start providing the forts to the east of Bearsden with annexes to house their replacement bathhouses, and there was also the Wall curtain to complete from Watling Lodg Carrideno et thin I . s last respec interestins i t i t noto gt e thasuperstructure th t annexe th f eo e rampar t Duntochea t havy f turo ma re t f bee(Robertsono f clad no yan n 1957, 63), thae th t superstructur annexe th f o ee rampar t Rouga t h Castle contained earth (Christison, Buchanan& Anderson 1905, 468) thad Wale an ,t th tha f l o tcurtai ease f Watlinth o t o nt g f eartLodgo s h ewa retaine clay db y cheeks. constructioPerhapw ne e sth n materials were being use r speede dfo du r o , shortageo t f suitablso ebuildine turth f fo towardd g programmeen e sth . It should not surprise us that it took so long to complete the Antonine frontier system in its final form. Wor continued kha Hadrian'n do s Wall frobeyono t mAc 2 D12 d AD 136. Indeed, there is good reason to believe that the true development of the Antonine Wall is far more complex than that laid out here. The decision to attach annexes to each fort, rather than to add a feature akin to the Vallum on Hadrian's Wall to the whole 38 mile (60 km) length of the Antonine Wall, would have resulted in a significant saving of labour during the construction phase, and would have eased manning problems thereafter. It might also have avoided difficulties in removing the indigenous populatioutilizatioe th n lane i th d df nan o thu r againss engulfedwa Chatte e th tth An n i i3 D (I .8 Domitia ordered nha d that compensation shoul made db r cropefo s thabeed ha tn includes hi n di : Frontinus Strategemata 11,11,7.) The suggested sequence also points us to the conclusion that the secondary forts at Bearsden and Duntocher were constructed at the same time as the Wall curtain. Had this not been the case, then they too could have been given annexes at a later stage, as was done with the primary forts built befor curtaie eth n (Gillam 1975).

USES OF ANNEXES Annexe e verar s y poorly understood e 196th 9n I .Archaeology editioe Th f no f Romano Britain they were noted as a special feature of the Antonine Wall forts, as being rare to the south of Hadrian's Wall, and as being used for civilian dwellings (Collingwood & Richmond 1969, 89). Such a view cannot now be held. Few Roman forts in Britain lacked an annexe as is clearly demonstrated by recent discoveries, especially those made through aerial reconnaissance (Wilson 1984; Maxwell & Wilson 1987). The problem rather has been one of over-concentration of archaeological investigation on the fort itself to the detriment of work on the fort's environs: aerial photographs have been usefu redressinn i l g this imbalance e pastth ,n I .annexe s have been discovered on the ground only during the excavation of a fort's ditch system, or where a surface scatte f artefacto r attractes ha s d interest generaa . Therw no ls ei recognitio n that annexes, like vici and other peripheral features, require further work. Even so, Anne Johnson's near-definitive book on Roman forts, published in 1983, fails to discuss annexes. accepe w equatioe f I th tmilitare th f no y zon f Hadrian'eo s Wall s defineVallume a , th y db , wit annexee hth e Antonin th f o s e Wall forts , e annexethe th e origina f ns certainlth o wa e s us l y military (Salway 1965, 156-8). Indeed defencee th ,annexe th f oftee o s e ar strons a n thoss ga f eo 306 SOCIET ANTIQUARIEF YO SCOTLANDF SO , 1994

ILLUS 3 Suggested building sequence for the Antonine Wall. The forts are numbered as follows: 1 Old Kilpatrick Duntocher2 , Castlehill3 , Bearsden4 , Balmuildy5 , Cadder6 , Kirkintilloch7 , 8 , r Hill Cro0 Ba 1 ,9 , y Hill Westerwood1 1 , Castlecary2 1 , Roug3 1 , h Castle4 1 , Falkirk, 15 Mumrills, 16 Inveravon, 17 Carriden, 18 BAILEY: THE PROVISION OF FORT-ANNEXES ON THE ANTONINE WALL I 307

e for whico th t t h the e intimatelar y y connected mosn i t te Antonin Ye .case th n o s e Walls a , elsewhere, they remained separated from the fort by a ditch, as befitted their secondary importance. That this separation was not always necessary is shown by the lack of ditches at Duntocher, e lasBearsdenth t n phasi d f Balmuildyeo an , e rampart Th annexe . th ford f o sean t appear to have been interconnected at Bearsden, although this may have been an accident of circumstance. All of the annexe ramparts on this frontier were attached to the Antonine Wall curtai hencd northera nan o et n cornee fortsth f .o r Unfortunately junctioe th e souther, th f no n ramparts of the annexes and those of the forts has never been clarified. At Duntocher it had already been destroyed before excavation (Robertson 1957, 64), wherea t Rougsa h Castlprobabls wa t ei y destroye fore th archaeologicaf m o y db l exploration retentioe Th inne.e th f rno ditche t Mumrillsa s and Castlecary doe t meano s n thaannexe th t e rampart t joino n e ford thost theith di a st f eo r southern corners. A parallel may be drawn with Gelligaer where the fort ditch was filled in at this point in its circuit to take the annexe rampart across it (Ward 1911, 80). However, even here the rampart then seems to have stopped 1.2 m short of the fort's own rampart, suggesting that there may have bee nwoodea n bridge-like structur patroe levee th th f t o lea walk annexe Th thu.s ewa s effectively an extension of the fort. annexew fe o s ss havA e been excavated eved an , n appreciably fewean n o r t e no scale s i t i , possible to attribute a definitive function or set of functions to them. They were not empty open spaces for they contained a large number of timber buildings. In general they are associated with semi-industrial activities which might have provided a health hazard had they been conducted adjacen e comfor barrackse th th o t f t o e tunit' Th . s bathhous s placee annexth ewa t i n di s ea probably presented a fire hazard. In the Falkirk and Inveravon annexes there is evidence, in the form of slag, for iron-smelting (Bailey forthcoming; Armit & Dunwell 1992, 179), and the large furnaces at Camelon may have serve dsimilaa r function (Maxfield 1979, 31)adjacenn A . tCameloe areth n ai n annexe showed thairoy an nt thus produced would have been reworke site smithinn th ei n do g hearths. Similar hearths, possibly associated with blacksmithing, have been found in the annexes at Falkirk (Bailey forthcoming b), at Mumrills (Robertson 1942, 119; Keppie & Walker 1989, 144), and the Flavian fort-annex t Elginhaugea h (Frere 1988, 429). Numerous example foun e alse b sar Englano on t di d (eg Melandra Castle: Goodburn 1978, 432) and in Wales (eg Gelligaer: Ward, 1911, 74). Eighteenth- and 19th-century ironstone mines are common in the vicinity of the Antonine Wall; they were presumably working the same mineral deposits as the Romans had exploited. Coal too was readily available and has been found in small quantities in most of the annexes of the Antonine Wall forts. Although there is no native lead available near the Antonine Wall it does occur in workable deposit s hinterlandit n i s ingon A .f leao t ds foun weighin wa associatio n di g k 8 g3 n with Roman potter e north-westh o fore yt t Cameloth a t f o t n durin constructioe gth e Midlanth f no d railway ther n 1848ei s foun secon A . same wa n 184 i dth g edn pi 9i area (Wilson 1863, 64) t weigheI . d s stampewa d dan IMP.CAES.HADRIANI.AVG.T.M.L.Vg k 2 10 s foun furtheA .wa t Kirkintillocha dg pi r . The Camelon pigs probably lay in an outer annexe where they may have been intended for reworking r havthey o , yma e been landee nearbth t da y harboumateriaw ra s r workina r fo l g sitee elsewhereth sf o wher e On e. this secondary processing could have taken plac s Rougei h Castle wher stonea e moul pewter d fo founs wa r d (Christison, Buchana Anderson& n 1905, 495). Copper, as well as lead, was worked in the annexes at Newstead (Jones, pers comm) and there are many instance f coppeso r workin annexen gi s from further south. s though time wa on t ei t t A that bric tild ekan manufactur s confine legionse wa e th o dt e th : work Legiocompoun X epitome X e seeth industrian b e t a s Holna a n th f eo f ca t do l annexe 308 SOCIETY OF ANTIQUARIES OF SCOTLAND, 1994

(Grimes 1930), with separate enclosure accommodatioe th r fo s e workforcth f no e paralleline gth auxiliare rolth f eo y fort detacheA . d enclosur fore t Mumrilleasfounth e s a t f th e o wa t o dt s (Steer 1961 tile , eth s nort 96) t kilit bu o ,nht mus connectee b t auxiliare th o dt y fort (Macdonald 1915). Confirmation of the role of auxiliaries in the manufacture of tiles is provided by the similar kiln at Gelligaer (Ward, 1913) and by the occurrence of tile stamps which name auxiliary units (Hassall 1979,264). Pottery productio alsn e considerenca o b activit n a e b yo dt undertake e auxiliarieth y nb s (Breeze 1986) discoveres kilA . nwa d r insidHil Ba fore l t e(Keppith a t e 1985, 60)anothed an , r outside that of Croy Hill (Hanson 1979, 20). Wasters indicate similar production at Duntocher (Gawthorpe 1980, 17), whilst fabric analysis suggests further local pottery manufacture at Inveresk (Thomas 1988), at Bearsden (Hird forthcoming), Westerwood (Webster forthcoming b) and at Falkirk (Webster forthcomin faro S ,. thougha) g , onle annex th kilya n i nt e Falkir a eb y kma associated with the manufacture of these. numeroue Th s pits found withi annexee nth s demonstrat neee eth discar o dt d large quantities of organic material. In general, inorganic waste does not require burial. Slag and burnt clay or potterusee b r surfacin dn fo yca g paths t CamelonA . r examplefo , whole th , e surfacpate on h f o e beed ha n mad frop eu m amphorae sherds (Maxfield 1979, 31)recene Th . t wor t Newsteaka s dha suggested that zonation was employed. Metal-working was carried out in the south annexe where over 10 strip-buildings have now been plotted; glass may have been worked to the east, whilst organic materials may have been processed to the west. These organic materials would have been quite varied: wood, bone, leather, animal organs, animal fats, glue, food-processing, textiles, etc. The army's demand for such organic provisions must have been immense; unfortunately they are t eas no detecarchaeologicae o yt th n i t l record. Leather-workin s beegha n deduce Welse th t da h site f Brithdio s r (White 1978 ) wher42 , hugea ecistera tand kan n were examined immediately outsid t Caernarfoa forte d eth an , n wher separatea e compoun 0 m 20 north-wes y dla fore th t f o t (White 1985). The evidence from the latter site included numerous leather off-cuts. In both instances these 'ordnance depots' were associated with other industrial activity: furnaces at Brithdir tila ed kilan , t Caernarfonna . Environmental evidenc t Vindolandaea e so-calleth n i , d fabrica, also indicates leather-workin auxiliarn a n gi y fort (Birley 1977, 123). Whil hidee eth f so the animals were used for leather the remainder of the carcass would have been fully utilized. Debris foun deea t outsidn i dppi ease eth t defence soute th f ho s cam t Camelopa n consistef do 'black sooty soil, containing broken potter collectiond yan f animaso l bones, teeth pieced an , f so deer horn. In many cases the bones formed layers, and were much decayed, while the teeth and horn were well preserved' (Christiso Buchanan& n 1901, 368) animale Th . s which providee dth raw materia processee mano s th r f yo fo l e annexsth havcarriey n i ema t e beedou n detainen di pens withi annexe nth e awaitin (Caruane gus a 1992). It would thus seem that the auxiliary soldiers stationed along the northern frontiers would have joined their more illustrious companions legionariese th , economie th n i , c exploitatioe th f no mineral resources of the area and of any raw materials furnished by the local population through taxation in kind. In this way they would have contributed to the supply of materials and products demanded in vast quantities by the Roman garrison of Britain (Breeze 1984a). The various specialists involve Romae th n di n arm wele yar l know t bearnbu repeating here. Ther s literarei y evidenc r farriersefo , master-builders, shipwrights, pilots, -makers, glaziers, smiths, arrow- makers, coppersmiths, helmet-makers, millers, wagon-makers, roof-tile-makers, swordcutters, water engineers, trumpet-makers, joiners, bow-makers, plumbers, blacksmiths, masons, lime- burners, woodcutters, charcoal-burners, butchers, huntsmen, grooms, tanners and cobblers, to whic mighe hw brewersd ad t , weaver dyerd an s s (Davies 1989, 40). While r somfo , e time, there BAILEY: THE PROVISION OF FORT-ANNEXES ON THE ANTONINE WALL 309 bees ha n abundant archaeological evidenc r thisonls i e fo t i y, wit discovere hth Vindolande th f yo a tablets thae havw t e written confirmation thaauxiliare th t y units containe da larg e par f thio t s diverse rang skillf eo s (Bowman 1994, 43-7). Many of these crafts did not require specialist buildings in which they could be undertaken. e open-endeTh d strip-building commonlo s , y neighbourhooe founth n i d f Romado n forts, would e adequateb e Flaviath t A .n for f Elginhaugo t h such buildings line e maidth n road througe hth annexe to the fort gate (Frere 1988, 429). In general, timber buildings seem to have been numerous within annexes, with considerable density in some instances. At Inveravon there were three phases of timber structures, each structure occurring quite close to its neighbour with narrow lanes fronting them (Armit & Dunwell 1992). A similar situation, with even more rebuilding, was recorde e soutth n hi d cam t Cameloa p n (Maxfiel e nortth d n h1979)i annexd an , e five superimposed timber-framed buildings were found (McCor Taid& t 1978, 156) larg.A e area containing sleeper beam trenches was also noted at Cadder during sand quarrying (Clarke 1933, 61). These appea havo t r e bee f severano l phases irregularite Th . f settinyo ford gan m that such buildings could take was demonstrated in the limited excavations within the annexe at Bearsden. The buildings, separated by cobbled areas, form no obvious pattern, p73 but do occupy much of the available space (Breeze 1984, 47). Overcrowding of the relatively small annexe at Balmuildy was given as one possible reason why the bathhouse had had to be constructed over the infilled fort ditches - not a desirable course of action (Miller 1922, 56). problee Th f identifyinmo g timber structure trian si l trenche clearls si y show t Duntochena r where occupation layers and pits were found in the annexe, but the associated buildings could not definee b d (Robertson 1957, 66)same Th . e proble encounteres mwa t Mumrillsda annexe th ; s ewa trenched by Macdonald who located post-holes, streets and a large 'boulder area' (Macdonald & Curie 1929), and trenched again by Steer who did find at least one timber building (Steer 1961, 93). By contrast, despite considerable post-Roman damage to the site, numerous timber structures were found in the open area excavation at Falkirk (Bailey forthcoming b). A variety of foundations were used, including sleeper walls of stone; post-holes; beam slots; and, probably, sill beams resting directly upon the ground. All the floor levels had been removed, and only fragmentary roatracee th df so syste m whic serviced hha buildinge dth s survived. Similar roads were founn di the earlier excavation annexee th f so t Mumrilla s s (Macdonal Curid& e 1929, 500), Rough Castle (Christison, Buchana Anderso& n n 1905, plan) Castlecard an , y (Christiso Buchanan& n 1903, 328), implyin presence gth considerablf eo e timbe f activito t no r f ybuildingi thesn si e areas. Two other types of buildings should be mentioned in connection with annexes. A wooden shrine or temple, with a tile roof, is evidenced at Balmuildy by the finding of sculptured figures of Victor Mard yan s (Miller 1922, 56)secone Th . d typ f buildineo knows gi mansio,a s na kina f do official guest-hous governmenr efo t personne memberd an limperiae th f so l post. Non thesf eo e ear known fro Antonine mth e Wall (unles hypocaustee sth d buildin t Kempega r Avenue, Falkirks wa , one e sout; th buildin n hi camV gX t Camelon pa , wit s opushit signinum floors should alse ob mentioned). littlo To e wor bees kha nannexese donth n eo , partly owinlace f structureth k o o gt s founn di the early excavations. However, the negative results have tended to come more recently from negative approaches. Trial trenches are not sufficient. Another problem has arisen in distinguishing between military annexes and civilian vici, a difficulty emphasized by Davies in Wales (Davies Thomay 1990b d an )Scotlann i s d (Thomas 1988). Bot annexn ha vicusa contain d ca an e n evidenc f industriaeo l activity, timber-framed buildings, large quantitie f Roman-stylo s e objects, and occupation phases which coincide with forte thosf equao th .f e eTheo b l proximitn yca e th o yt neitheford an t r needregularle b o st y planned dispute Th . e ove firse rth t settlemenphase th f eo o t 310 SOCIET ANTIQUARIEF YO SCOTLANDF SO , 1994

the west of the fort at is a good illustration of the problem. It has a regular plan which incorporate fore sth t bathhouse possibla , e mansio, copper-workina d an g centre. Furthermoree ,th whol f thie o surrounde s si larga y db e clay rampart (Birley 1977 ,f thi o 70)consisten s sl i .Al t with an annexe and broadly similar arrangements can be found at Melandra Castle (Wilson 1970, 285) and Prestatyn (Frere 1986, 365) lattee .Th bathhousra sits eha e withi defendena d enclosured an , evidence of copper-working in two of the six known buildings. (Moulds for brooches, crucibles, hearth linings and scrap bronze were found.) Yet the excavator at Vindolanda has identified phase I as a vicus, and indeed it is followed on the same site by an unenclosed settlement of less regular plan. Similar confusion occurre t Castleforda d wher e areeth a aroun e largdth e stone bathhouse was initially interpreted as a vicus, but has subsequently been called an annexe - an all too common mistake. At Slack and Bainbridge claims have also been made for defended vici (Hartley & Fitts 1988, 65). (A good discussion of the possible status of annexe-type vici is given with reference to the defended enclosures around the special site of Ribchester: Olivier 1987.) Sommer's solution to this particular problem was to adopt the somewhat ambiguous term 'military vicus' for these enclosures and to postulate the existence of a class of seconded civilians, sometimes know s 'camna p followers' whicn i ; h cas t woulei d have been these quasi-military personnel, protecte e militarth y db y whom they served wero wh ,e responsibl r supplyinefo e gth army's needs (Sommer 1984). It does seem, therefore, that forts with annexes were probably the norm, and that they were multifunctional with specialized production zones. In this case we ought to be questioning the usual identification of one of the buildings in the central range of forts as a fabrica: the regimental workshop f theso w e buildingFe . s have produced direct evidenc manufacture th f eo e of equipment and most can at best be regarded as repair shops (Johnson 1983, 186). Nor are these buildings sufficient in size to cope with the regimental demands for tools or weapons. How r examplefo , , coul singlda e building, less than one-thir centrae aree th f th ao f ldo rangf eo a fort, hold the 343 men (12 of them shoemakers) who are attested at the workshops at Vindolanda (Bowman 1994, 106)? Their identification as fabricae rests largely upon the analogy with the layout of the ideal temporary camp described by Hyginus, and upon excavated examples within legionary fortresses. Neithe f theso r e situation directle b n sca y compared with auxiliare th y fort. Indeed t wouli , d see fora m havo r t t rathefo elarga d od re defended annexe specificall locato t y e such industrial activities thed placo an ,t n a solitare y workshoe th n pi e centrforth t f amongso e e administrativth t livind an e g quarters .e excavateMosth f o t d buildings that have been called fabricae would have been much better suited to the function of storage of the finished articles rather than to their production. This would make sense, as the mora s forewa tsecur e locatio sucd nan h supplies woul e neededb largn di e quantities should the garrison find itself besieged. Items such as tools, weapons, armour, textiles (blankets and uniforms) and pottery could have been stored separately and this may explain why many of these buildings have subdivisions. Amongs weapone th t s e ballistaewoulth e db o whict d hha be kept dry if they were to be of any use. Several building inscriptions in the north of Britain poin suco t t h covered store r catapultfo s s (ballistarium) alsd an o indicate that they occupied centrae th par f o tl range (Donaldson 1990, 211). Armamentaria (general weapon storer so magazines) are mentioned in several inscriptions from within forts. That from Lancaster (RIB 1092) records that 'principia et armamentaria conlapsa restituit' (restored the derelict principia and armamentaria); this, together with the finding of military equipment in and around the principia of other forts has led to the identification of the courtyard rooms of the principia as the pertinent stores (Johnson 1983, 108). However wordine th ,Lancastee th f go r inscription indicates tha a separatt e structur havy ma ee been involved n armamentariuma d an , s i BAILEY: THE PROVISION OF FORT-ANNEXES ON THE ANTONINE WALL 311

m C 0 5 10 15 20

I I 50

CROY HILL

IE

D KILPATRICOL K ILLUS 4 Plans of the 'granary' at Croy Hill and building XII at Old Kilpatrick

mentioned on its own at Leiden-Roomburg (Bishop & Coulston 1993, 199). The building adjacen principiae th o t t withi centrae nth l range would therefor quite eb e appropriate. Indeed, t identifno d di y e thesiw f e building s suc sa woule hw troublee db explaio dt sucy nhwh stores were absent.

ACKNOWLEDGEMENTS I would lik thano et k David Breeze, Gordon Maxwel Richard an l d Hingle commentinr yfo a n go draft of this article. The Walnuts have lived with the theories herein advocated and have helped with their gestation. Errors remain my own.

NOTES 1 The buttressed building to the north of the principia at Croy Hill is usually interpreted as a granary. However, there are a number of problems with this. It is only 14.3 m long externally and it went out of use long beforfort e traceo th rese N . f eth f sleepeo t so r walls were interioe founth n di r although there was (unusually) a drain, and there was possibly a lateral entrance near the south-west corner. Unfortunately t knowno s i n t i ,whethe r there were granarie laterae th n si praetorii, thoug excavatoe hth r thought this probable (Macdonald 1936, 60)buildine Th . g resemble internae sth l bathhous t Mumrillea s remainy (c, an e lacexcephypocausa f4) f th k f o illur o s& fo t2 s t system e buttresseTh .e th n o s bathhouse at Mumrills can also be found at Camelon on buildings 17 and 18 (Christison & Buchanan 1901, 371), usually interprete bathhousess da firse th t n bathhouso d an , t Bearsdenea . 2 Like the internal bathhouses mentioned, the building in note 1 went out of use before the end of the fort and was cobbled over to create an open square in front of the principia. An external bathhouse is known from this site, but insufficient is known of it to be certain that it was a replacement. 3 Again the interpretation of a stone building within a fort is difficult, owing to the degree of demolition that occurred in the Roman period. The size of the structure at Old Kilpatrick is indicative of a bathhouse and there is a gap in the foundation of the original cell in the centre of the north wall which could have 312 SOCIETY OF ANTIQUARIES OF SCOTLAND, 1994

served as a flue. With the addition of another room to the south a similar gap occurs in the foundation wall betweeroomso tw e .nth Thihavy ma se bee communicatinna g hot-air channel thirA . d passage occursoute th n hadditione si walth f o le sub-floo Th . r containe traco latrina dn f eo e channel arouns dit walls whole th ; e basemen beed ha tn fille witn di h rubbis rubbld han e onc buildine eth f gono d t gha eou use (Miller 1928, 28). This suggests that the floor may have been supported on pillars which have subsequently been remove e elsewhereus r dfo . Lik e otheth e r known internal bathhousese th d an , structure at Croy Hill, this building was also cobbled over.

REFERENCES Armit, I & Dunwell, A 1992 Excavations on the Antonine Wall Fort at Inveravon, near Grangemouth. Edinburgh. Bailey forthcominB G , a g'Th e Antonine frontie Callendan i r r Park, Falkirk s forstructurad it ; man l sequence'. Bailey, G B forthcoming b 'The Roman Fort on the Antonine Wall at Falkirk'. Bailey, G B & Keppie, L J F forthcoming 'Recent work on the Antonine Wall'. Birley, R 1977 Vindolanda: A Roman Frontier Post on Hadrian's Wall. London. BishopCoulston& C N 199 M , C J , 3 Roman Military Equipment. London. Boon (edC G ), 1978 Roman Sites. Cardif Cambria= ( f n Archeol Assoc Monogr. 1) , Bowman, A K 1994 Life and Letters on the Roman Frontier. London. Breeze, D J 1975 'The abandonment of the Antonine Wall: its date and implications', Scott Archaeol Forum, 7(1975), 67-80. Breeze, D J (ed) 1979 Roman Scotland: Some Recent Excavations. Edinburgh. Breeze J 1984D , a 'Deman supplnortherd e dth an n yo n frontier' Miketn i , Burgess& R , (eds)C , , Between and Beyond the Walls: Essays on the Prehistory and History of North Britain in Honour of George Jobey, Edinburgh, 264-86. Breeze, D J 1984b 'The Roman fort on the Antonine Wall at Bearsden', in Breeze, D J (ed), Studies in Scottish Antiquity presented Stewano t Cruden, Edinburgh, 32-68. Breeze, D J 1986 'The manufacture of pottery in Roman Scotland', Proc Soc Antiq Scot, 116 (1986), 185-9. Breeze, D J & Dobson, B 1970 'The development of the mural frontier in Britain from Hadrian to Caracalla', Proc Soc Antiq Scot, 102 (1969-70), 109-21. Breeze Dobson& J D , B ,198 7 Hadrian's Wall, 3 edn. London. Burnham, B C & Davies, J L (eds) 1990 Conquest, Co-existence and Change: Recent Work in Roman Wales. Lampete Trivium,= r( 25). Caruana D 199I , 2 '; excavatio sectioa annexe f th no f no e firs e ditcth tf hFlaviao n fort, 1990', , 23 (1990), 45-109. Christison, D & Buchanan, M 1901 'Account of the excavation of the Roman station of Camelon, near Falkirk, Stirlingshire', Proc Soc Antiq Scot, 35 (1900-1), 329^117. Christison, D & Buchanan, M 1903 'Excavation of Castlecary fort on the Vallum', Proc Soc Antiq Scot, 37 (1902-3), 271-346. Christison, D, Buchanan, M & Anderson, J 1905 'Report on the Society's excavation of Rough Castle on the Antonine Vallum', Proc Soc Antiq Scot, 39 (1904-5), 442-99. Clarke J , 193 Romane 3Th Fort t Caddera (near ). Glasgow. Collingwood Richmond& G R , I , 196 Archaeologye 9Th of . London. Daniels, C M 1978 Handbook to the Roman Wall with the Cumbrian Coast and Outpost Forts, 13 edn. Newcastle. Davies L 199J , 0 'Military Vici: recent researc s significance'it d han Burnhamn i , Davies& C L B J ,, (eds), 1990. Davies , Breeze MaxfieldW & J R , D (eds, A V , ) 1989 Service Romane th n i Army. Edinburgh. Donaldson H 199 G , 0 'A reinterpretatio 191B RI 2 f frono m Birdoswald', Britannia, (1990)1 2 , 207-14. Frere S 198S , 6 'Roman Britai 1985n ni ; sites explored', Britannia, (1986)7 1 , 364-427. BAILEY: THE PROVISION OF FORT-ANNEXES ON THE ANTONINE WALL 313

Frere S 198S , 8 'Roman Britai 1987n ni ; sites explored', Britannia, (1988)9 1 , 416-84. WilkesFrere& S J S ,J 198, 9 Strageath: Excavations within Romane th Fort 1973-86. Londo= n( Britannia Monogr Ser, 91). Gawthorpe, S 1980 'Roman pottery from the Antonine fort and fortlet, Golden Hill, Duntocher', ClydebankHistJ, (1980)1 , 15-18. Gentry P 197A , 6 Roman Military Stone-built Granaries Britain.n i Oxfor Bri= d( t Archaeol Rep, 32). Gillam P 197J , 5 'Possible change course constructioe plan th th i s n f ni eo Antonine th f no e Wall', Scott Archaeol Forum, 1 (1975), 51-6. Goodburn R ,197 8 'Roman Britai 1977n ni ; sites explored', Britannia, (1978)9 , 404-72. Grimes, W F 1930 Holt, Denbighshire: The Works Depot of the Twentieth Legion at Castle Lyons. London (Cymmrodo=Y r 41). Hanson S 197W , 9 'Croy Hill' Breezen i , (ed)J D , , 1979,19-20. Hanson, W S & Maxwell, G S 1983 Rome's North-west Frontier: The Antonine Wall. Edinburgh. Hartley, B R & Fitts, L 1988 The . Bath. HassallC 197 W 9M , 'Military tile stamps from Britain' McWhirrn i , (edD )A , 1979. Hird forthcominL , g 'The coarse Pottery RomanBreezee n i ' Th J D ,Fort t Bearsden.a Jarrett G 196 M , Romane 9Th Frontier Wales,n i edn2 . Cardiff. Johnson A ,198 3 Roman Forts. London. Keppie F 198J L , 5 'Excavation Romae th t a s r Hilln ForBa ,f o t1978-1982' , Glasgow Archaeol2 1 , J (1985),49-81. Keppie, L J F & Murray, J F 1981 'A Roman hypocausted building at Falkirk', Proc Soc Antiq Scot, 111 (1981), 248-62. Keppie, L J F & Walker, J J 1989 'Some excavation along the line of the Antonine Wall, 1981-1985', Proc Soc Antiq Scot, 119(1989), 143-59. McCord Tait& J N ,, 1978 'Excavation e northerth n i s n Annex Romae th f o en For t Camelona t , near Falkirk, 1961-3', Proc Antiqc So Scot, 109(1977-8), 151-65. Macdonald G , 1915 'Some recent discoverie Antonine th line f th eo n so e Wall', Proc Antiqc So Scot,9 4 (1914-15) 1383 9 , . Macdonald G ,193 2 'NoteRomae th n so nKilpatricd fortOl t a s Crod kan y Hill', Proc Antiqc So Scot,6 6 (1931-2), 219-76. Macdonald G ,193 3 'NoteRomae th n so n Fort t Rougsa h Castl Westerwoodd ean , wit hpostscript'a , Proc Soc Antiq Scot, 67 (1932-3), 243-96. Macdonald, G 1934 The Roman Wall in Scotland, 2 edn. Oxford. Macdonald G , 1936 'A furthe Romae r notth n eo n for t Croa t y Hill', Proc Antiqc So Scot, (1935-6)1 7 , 32-71. Macdonald, G & Curie, A O 1929 'The Roman Fort at Mumrills near Falkirk', Proc Soc Antiq Scot, 63 (1928-9), 396-575. MacMullen, R 1967 Soldier and Civilian in the Later . Harvard. McWhirr, A D (ed) 1979 Roman Brick and Tile. Oxford (= Brit Archaeol Rep, S 68). Maxfield, V A 1979 'Camelon 'South Camp': excavations 1975-79' in Breeze, D J (ed) 1979, 28-32. Maxwell, G S 1972 'Excavation at the Roman Fort of Crawford, Lanarkshire', Proc Soc Antiq Scot, 104 (1971-2), 147-200. Maxwell, G S & Wilson, D R 1987 'Air reconnaissance in Roman Britain 1977-84', Britannia, 18 (1987), 1-48. Miller N 192S ,Romane Th 2 Fort t Balmuildya (Sumerston, near Glasgow) Antoninee th n o Wall. Glasgow. Miller, S N 1928 The Roman Fort at Old Kilpatrick. Glasgow. Olivier, A C H 1987 'Postscript: the nature of the Ribchester civil settlement', in Edwards, B J N & Webster (edsV P ,) Ribchester Excavations, Part : Excavations2 e Civilth n i Settlement 1969-1980, Cardiff, 117-26. Owen-John S 199H , 0 'Recent excavation Romae th t a s n auxiliary fort t Loughoa s Neathd an r , West Glamorgan' Burnhamn i , Davies& C (ed)L B , J , , 1990, 40-8. CollinB RI g Wright wood& G P Roman196e R , R , Th 4 Inscriptions of Britain 1. Oxford. 314 SOCIETY OF ANTIQUARIES OF SCOTLAND, 1994

Robertson, A S 1942 'A Roman oven at Mumrills, Falkirk', Proc Soc Antiq Scot, 76 (1941-2), 119-27. Robertson, A S 1957 An Antonine Fort: Golden Hill, Duntocher. Edinburgh. Salway P , 196 Frontiere 5Th People of Roman Britain. Cambridge. Sommer A 198 C ,Military e 4Th Vici Romann i Britain. Oxfor Bri= d( t Archaeol Rep, 129). Speidel P 198M , 7 'The Chattan War Brigantiae th , Antonine n losth e Revol f th so d ean t Wall', Britannia, 18(1987), 233-7. Steer A 196 K , 1 'Excavation Mumrillt sa s Roman fort 1958-60', Proc Antiq Scot, (1960-61)4 9 , 86-132. Steer A 196 K , 4 'John HorsleAntoninthe yand e Wall', Archaeol Aeliana, ser4 (1964),42 , 1-39. Thomas, G D 1988 'Excavations at the Roman civil settlement at Inveresk, 1976-77', Proc Soc Antiq Scot, 118(1988), 139-76. Ward, J 1909 The Roman Fort of Gelligaer: The Baths', Trans Cardiff Naturalists' Society, 42. Ward, J 1911 'The Roman Fort of Gelligaer: The Annexe', Trans Cardiff Naturalists' Society, 44. Ward, J 1913 'The Roman Fort of Gelligaer: Discoveries made in 1913', Trans Cardiff Naturalists' Society,. 46 Webster, P V forthcoming a 'The pottery' in Bailey, G B forthcoming. Webster, P V forthcoming b 'The pottery' in Keppie, L J F forthcoming The Roman Fort at Westerwood. White, R B 1978 'Excavations at Brithdir, near Dolgellau 1974' in Boon, G C (ed) 1978, 35-62. p73 White B 198 R , 5 'Excavation Caernarfot sa n 1976-77', Archaeol Cambrensis, (1985)4 13 , 53-105. Wilson D ,186 3 Prehistoric Annals of Scotland, . Edinburghed d 2n . Wilson, D R 1970 'Roman Britain in 1969: sites explored', Britannia, 1 (1970), 269-304. Wilson, D R 1984 'Defensive outworks of Roman forts in Britain', Britannia, 15 (1984), 51-61.