In Wales and the border region we are lucky in having had so much interest shown from castellologists over the years, and such a tradition still continues to thrive in this interesting and stimulating region.  With all the available evidence, where is it best to commence this discussion?  Probably it is best to start at known points and it must be admitted that the only area with really firm dating evidence is the Edwardian Conquest castles of North Wales.  However even here our knowledge of the earlier tenurial history of these sites, viz Caernarfon, Buellt and probably Aberystwyth (3 earlier castles at this site?) is sadly lacking.  Caernarfon and Buellt have a history dating back to the earliest Norman invasions and Aberystwyth to only slightly later.  Little research has been done on the early history of all three sites, especially the first two.  It is possible that in Caernarfon this will have a leaning on the study of the masonry remains.  Certainly Victorian photographs of the interior of the castle show earthworks which have now been swept away by tidying up operations.  My own research at Builth Wells shows there is good evidence that the present castle site dates only from 1210 and not the eleventh century as previously believed.

From the plans of these royal Edwardian castles, it is generally assumed that the baronial houses spawned their own comparable designs, viz Holt, Caergwrle (Hope), Ruthin, Denbigh and Chirk.  But did the lords of these castles knowingly copy their king's designs?  Was Roger Mortimer Junior influenced by the North Welsh castles or in fact was Beaumaris based on Chirk?  And what of Caerphilly (pre-dates all Edward’s castles) and Morlais, the pinnacles of baronial castle building?  A point in question may be Flint.  If this had been more thoroughly ruined than it currently is - say all the walls and towers had now gone except for the two southern walls of the baileys with the ruined entrances still set in them as they are now - what would be left?  A turfed over mound of rubble or 'motte' where the keep had been (many turfed over collapsed keeps have in fact been mistaken for mottes, cf. Richards Castle, Dinas Emrys etc.), and two baileys which show little, if any, evidence of being flanked and had just simple hole-in-the-wall gateways.  Who would not by current standards, on architectural grounds alone, date such a structure to the twelfth century?  Without tenurial history and a complete ground plan (earthworks, not just masonry), preferably buttressed by excavation, any dating attempt is likely to be flawed.  Even with the increasing abundance of documents concerned with castle building in the thirteenth century, it is rarely possible to be conclusive about a castle's origin.  Yet we are prepared to date military structures to narrow, pre-conceived, non-contemporary bands, nearly always in total isolation from the tenurial history of the site.

If our grasp of the architecture of the royal Edwardian castles of Wales is uncertain and incomplete, it is to be expected that our knowledge of less well-known sites is even less secure.  Much useful work has been done on the career of William Fitz Osbern in Herefordshire and the border, but detailed work on his known and possible castles is generally lacking.  The building of the hall at Chepstow is generally attributed to him or his son, Roger, before 1075, but proof of this is lacking.  It is widely accepted, with the exception of Ludlow, that in Wales and the Marches between the dispossession of Roger of Breteuil (1075) and the reign of Henry II (1154-89), no masonry castles were constructed.  In my opinion this is largely due to a failure to take into account the tenurial history of castle sites.  It is therefore necessary here to look at some of the available history for these fortresses.  To keep this paper to a reasonable length I will attempt to limit this discussion to round towers.

The standard view for round towers is that they are a military feature of the thirteenth century.  This view is based on a handful of 'known' building dates which in themselves are not totally satisfactory, viz Beeston (SJ.537.593) 'circa' 1220-25 and Bolingbroke (TF.349.649 - 'early history unsatisfactory' - D.J. Cathcart-King) 'circa' 1220-32.  Similarly Welsh style D-shaped towers are classified according to what little is known about Castell y Bere, allegedly commenced in 1221 as Llywelyn's castle in Meirionydd.  In reality the earliest evidence we have for round towers in Wales is the event that occurred almost certainly in the winter of 1165/6, retold by Giraldus, concerning the death of Mahel, lord of Brecon.  During the night Bronllys castle caught fire and whilst fleeing the holocaust a stone fell from the 'principal tower' on to Mahel's head.  The linking of the tower on the motte to this event is currently rejected on architectural and no other grounds!  Instead the current tower is said to have been built by Walter Clifford (1207-68).  Of him we have no evidence that he built or was even interested in building anything.  He is accredited with the conception of this tower simply because he was the owner and was living 'circa' 1220-30 when it is claimed round towers were being built^*[1]!  The motte on which the tower stands shows evidence (bedrock) of being natural and must have been a primary feature of the site and therefore the castle's principal tower was most likely built upon this from the castle's conception, probably in 1144.  The tower also shows evidence of burning and of having been altered internally several times, as well as of being heightened.  Remains of a stone forebuilding can still be made out leading up to the entrance doorway from the bailey.  If this tower is built on bedrock then there is no reason for it not having been built from the first - once again no need for settling of the 'motte' - and there is no reason to doubt that the windows could have been later insertions.  Comparisons with similar round towers on mottes strengthens this argument.  Since this was written excavation has proved that the tower is in fact based on an earlier polygonal tower or possibly a small shell keep.

Buellt or Builth Wells castle, as it is now known, was destroyed in 1169 by Rhys ap Gruffydd of Deheubarth and was later 'constructed' for King John in 1210^*[2].  The site has always been reckoned to be the site under the Edwardian castle.  However just west of the current town is another motte and bailey, and on the motte are the buried remains of a round tower equivalent to those found at Bronllys and Longtown.  This probable early tower-keep is quite likely the first Braose castle of Buellt, founded in 1093.  This immediately brings into question Longtown's alleged date.  Longtown keep has generally been dated to the thirteenth century.  However in this case the historic dating evidence does point clearly (but of course without total certitude) to the period immediately prior to 1230, even though the castle appears to have been newly commenced in stone in the 1180s^*[3].  The conclusion that the keep was built at this time from the dismantled remains of whatever preceded it is re-enforced by the re-used material in the round keep.  This 'evidence' however, can be argued against on the sensible grounds that it is not conclusive.  So too can the dating of any structure as I think has been adequately stated above. 

The above sites beg the question as to what conclusive evidence there is for twelfth century (and possibly older) round towers?  One castle springs immediately to mind - Nevern in Cemais (SN.083.402).  Here are the barely noticed remains of a round keep set in a motte; and although we have no building date for this structure we do have a destruction date - 1195!  The castle itself was probably founded soon after the occupation of Ceredigion in the reign of Henry I (1100-35), and it is possible that the keep was built soon afterwards on terra firma, as has been proved at Skenfrith and suggested at Clifford and Bronllys.  According to an unfortunately lost DOE excavation report, Longtown was built on the motte on the most insecure of foundations.  In the other cases the mound was subsequently thrown up around the base of the keep, which was built on a solid foundation.  There is therefore no need to wait for the soil above terra firma to settle.

It is also worth noting here what happens when the upper stories of such a keep are removed.  You are left with a mound, possibly with rubble surrounding it or vaguely protruding from it.  Such mottes/mounds which have been excavated revealing a masonry keep and are not mentioned in this text include Barnstaple (SS.556.334), Crug Eryr (SO.158.593), Dolforwyn (SO.153.950), Elmley (SO.980.403), Duffield? (SK.343.441), Kenfig (SS.801.827), Miserden (SO.944.093), Neroche (ST.272.158), St Briavels (SO.558.046), Castell Dinas Emrys^*[4], and Winforton (SO.302.463, probably not defensive^*[5]).  Those that show definite evidence of once having supported or having been towers include Bleddfa (SO.209.682), Caus (SJ.338.079), Chartley (SK.010.825), Clun (SO.298.809), Crickhowell (SO.218.183), Dolwyddelan I (SH.725.522), Lyonshall (SO.331.563), Mileham (TF.312643), Plympton (SX.546.559) and the small shell keep at Wiston (SN.022.181).  Those that show some evidence include Aberedw (SO.078.471), Dolbenmaen (SH.506.430), Eardisley (SO.311.491), Huntington (SO.249.539), several of the castles of the Radnor plain^*[6], Old Castleton (SO.283.457), Ongar (TL.554031), Oswestry (SJ.290298), Pilleth (SO.259.677) and Tomen y Mur (SH.705.386).  This list is by no means definitive and no attempt has been made to list the sites chronologically.

There are other such round keeps in Wales that may well be twelfth century.  Probably the best such case is the keep at Dinefwr.  A 'tower' here was strongly defended in 1213 after the rest of the castle had fallen^*[7].  Such a tower is only likely to have been the present round tower.  The 'tower' at Llandovery successfully resisted attack after the fall of the bailey in 1116.  Once again the castle remains here have not been properly surveyed and, considering the tenurial history, in my opinion are most unlikely to date to 1282 and after.  The castle of Rhaeadr-gwy in Gwrtheyrnion is another good example of a historically recorded, but now destroyed stone tower on a 'motte'.  Built by Rhys ap Gruffydd in 1177 it was twice destroyed by the princes of Maelienydd, the last occasion in 1194 proving final.  Here is an attested Welsh built stone tower with a terminus post quem and a terminus ante quem.  It is unfortunate that so little of the structure now remains^*[8].  Regardless of this, time and time again, round towers are dated to the thirteenth century on no grounds other than vague architectural assertions, viz. Dolbadarn, Pembroke, Cardigan, Usk, Penrice, Wigmore, Clifford, Bryn Amlwg, etc.  What actual firm evidence do we have for the building of these structures?  Can we really date them on the grounds of architectural features, the construction of all of which undoubtedly spanned many generations?  There can be no hard and fast rule.  Of all the historical documentation that may have existed about a castle site we probably are lucky if we can now muster 1% of it.  The whole of that documentation probably covered but 1% of what we would now want to know about that site.  Can we therefore really be so certain that Hay castle was built c.1200 by Matilda Braose, Bronllys tower was built c.1220 by Walter Clifford or even that Clifford castle was built c.1220 by the same much put upon Walter Clifford?  Merely on the evidence of windows which may have been replaced, most likely repeatedly, and styles that were neither fixed nor compulsory?

The case of Nevern (above) opens interesting speculations.  It is generally argued that the chief lordship castle was copied by the lesser baronial castles.  In many cases this is demonstrably false, though this is not to say that it did not happen.  If Nevern was a copy of the lordship round tower at Cardigan, then Cardigan's tower is logically older.  The structure there is certainly unusual.  It is said to be a stepped round tower with deeply projecting corner spurs, making the ground plan square, rather than round, though this may well be a confusion between the 'great tower' and the 'watchtower'.  Were either of these built in the attested stone building of the castle by Rhys ap Gruffydd in 1171?  Was the tower already there then, or was it the great tower whose second storey was constructed for Henry III in 1261^*[9]?  Even though we have a historical mention of building work at a castle, this cannot be taken as evidence that this involves the erection of whatever piece of masonry we are currently interested in.  Any castle that has been adequately examined shows multiple building phases, and these indeed are to be expected over any length of time.  It is not so much building evidence that we lack, but a sensible method of relating this to the castle.  Currently not enough research has been undertaken on the tower at Cardigan to even hazard a guess as to its construction date or its repair.  The same should be true of Pembroke, for here we currently have no historical help^*[10].  Giraldus mentions Pembroke castle, initially hastily and roughly constructed in the form of a stockade.  From this point he tells us stories of the siege of 1096, but of the later structure, which must have followed the rough and ready stockade he tells us nothing.  We know that the site was of great strategic importance during the twelfth century, yet it is not until "c.1200" when once again all the masonry was suddenly supposed to have been built, even if the Old Hall is allowed an unconvincing date of "c.1180".  All of this architectural dating evidence is quoted despite the fact that Gerald Windsor, the castellan, surrounded his new castle, which was probably at Cilgerran, with a wall as long ago as 1108^*[11]!  Indeed evidence of this apparently 'clay-laid' wall^*[12] still remains under the later gatehouse and on the rock ledge beneath the castle's current north curtain wall.  If Cilgerran was stone-built from the first then what of the myriad of other secondary castles that litter Wales and the border?  There is evidence of long abandoned stonework at so many!  Again the remains of the two round towers and linking curtain at Cilgerran are dated to after 1223 on the grounds that fighting was going on then and a re-fortification (refirmavit), not a complete rebuilding of a derelict site, was attested^*[13].  However it must be stated that the castle had been the scene of the bitterest fighting since the 1160's and had been frequently taken and rebuilt^*[14].  The two giant round towers are also obviously of different builds, one probably dating to the 1220's rebuilding the other on stylistic grounds obviously being much older.

It is also a long held view that a powerful ruler was more likely to build a substantial castle than a weak one.  Hence the attachment of many masonry castles to well-known personalities viz. Richard of Cornwall, William Marshall, Walter Clifford etc.  Yet how much weight can really be attached to these fleeting arguments?  Was a lord who is now seen as great really more likely to build a masonry castle than one who now appears insignificant - insignificant being directly related to short-lived, militarily unadventurous (he might have preferred building castles rather than fighting battles) or simply unrecorded?  Roger Mortimer of Wigmore (1246-82) was undoubtedly the most powerful member of that family in the thirteenth century, yet it was his little-known father Ralph (1226-46) who was responsible for founding/strengthening Knucklas and Cefnllys castles and probably several others!  Death also struck at odd times and many a promising career was cut short by the hand of fate.  In how many other cases may a short-lived powerful and energetic lord be dismissed as irrelevant simply because their period of lordship did not span forty years?  Our own political leaders rarely last ten and look what building works some of these have achieved!

Powerful castles fell to an enemy through surprise or lack of will to resist.  Rarely did they fall to determined attack.  There are few well chronicled sieges in the Middle Ages.  One of these is concerned with the siege of the powerful stone castle of Bedford in 1224^*[15].  Bar the motte nothing at all can now be seen of this castle, yet we know it was once one of the most powerful castles in the kingdom, with two stone walled baileys and a round tower set in a round shell keep on a motte, of which we even have a contemporary, if stylized, drawing by Matthew Paris [compare this to the drawing of Lincoln castle during the battle of 1217].  We also know that it took the king two months and all the power of the kingdom to subdue this determinedly held castle.  The first thing to be said about Bedford is that it was too strong for King Stephen to take in 1138 and that he had to resort to starving the garrison out.  This suggests stone defences.  When Faulkes Breauté gained the castle in 1215/6 he proceeded to strengthen it with 'towers and outer defences' and 'towers and walls and outer walls from the stones of the churches, and surround it on all sides with deep moats of pavement'.  Deep moats of pavement would suggest strong reveting of the ditches due to the low lying nature of this probably wet site.  It is worth quickly quoting here the main occurrences of the siege by Henry III which began on 20 June 1224.  The siege began with the outer ward being battered with mangonels and petraries.  Subsequently the barbican was taken, then breaches were made in the wall of the outer bailey and this was next stormed with much loss of life.  In return for their sacrifice most of the garrison ammunition, horses and livestock fell to the attackers.  The king then ordered belfries constructed and whilst his archers maintained a murderous fire on all that moved in the castle, a tortoise breached the inner curtain by the 'ancient tower^*[16]'.  Again this ward was then successfully stormed with great loss of life.  It was now necessary to undermine the keep on the motte and about Vespers on 14 August the wooden pit-props were fired and great gaps appeared in the shell(?) wall.  The garrison then surrendered to the king's mercy and Henry ordered the castle demolished.  Despite the protests of the owner, William Beauchamp, the keep and the outer bailey were totally demolished and the inner bailey was reduced to half its former height.  So ended one great round keep!  In Speed's day only the mound and 'old ruins' of the inner ward remained, attesting the thoroughness of the destruction and the accuracy of the accounts!  From this we can see that in 1224 a massive stone castle stood in Bedford and that parts of its masonry already dated from 'antiquity'.  Other masonry castles must have been of a similar age while the round keep at New Buckenham, Norfolk, is generally accepted as dating to a few years either side of 1140^*[17].  Finally of course there is the indisputable fact that rectangular towers were constructed throughout the Middle Ages, cf. those built at Harlech in 1323-4^*[18].  If this is true of rectangular towers why then should round towers have been any different?