William Marshal, Pembroke Castle and Angevin design Neil ...

213THE CASTLE STUDIES GROUP JOURNAL NO 29: 2015-16THE CASTLE STUDIES GROUP JOURNAL NO 32: 2018-19

William Marshal, Pembroke Castle and Angevin designNeil Ludlow

William Marshal, Pembroke Castle and Angevin design

Pem

brok

e C

astle

kee

p 12

01-1

207.

Vie

w fr

om th

e so

uthe

ast.

© N

eil G

uy


William Marshal, Pembroke Castleand Angevin design

This paper argues that the castles of Wil-liam Marshal the elder owe their influencesalmost entirely to the works of the EnglishCrown – as might be expected given hisclose association with the Angevin kings.The dominant narrative, suggesting influ-ence from the court of the French kingPhilip II Augustus, fits in with neither form,function, detail and relative chronology,nor with the political circumstances of theperiod. Philippienne influence only be-comes apparent in the work of his sons,after 1219. Attention will accordingly bedrawn to the cylindrical donjons built inFrance by the Angevins, and to the distinc-tion between the ‘donjon’ and other formsof dominant tower. By 1201, when work at Pembroke Castlebegan, Marshal had spent over fifty years inand around castles on both sides of theChannel, as squire, sportsman, soldier andcourtier. He had plenty of opportunity tofollow their development in the Angevindominions and beyond. His experience ofthe French royal domain was howeverlargely confined to the pre-Philippienneperiod. He is recorded in the Île-de-Francejust four times after Philip’s succession in1180, and each visit was short, and circum-scribed. There are several published accounts ofPembroke Castle (e.g. King 1977; King1978; Renn 1968), and descriptions hereinwill be brief. See Appendix 1 for a discus-sion of the function of its donjon.The cylindrical donjons of the Angevinkings

Excavation suggests that the origins of thefree-standing cylindrical donjon lie in theLoire region, at Fréteval (Loir-et-Cher,

Blois), built very early in the twelfth cen-tury by an ally of the Angevin counts(Mesqui 2013, 185 et al.).1 The form wasadopted by the counts of Blois during thetwelfth century, and in Angevin-held Tou-raine, Maine and Normandy. Of the 52 cylindrical great towers inFrance that were begun before c. 1203(Table 1; Fig. 1),2 nearly half (23) arethought to have been built by the Angevinkings or their vassals. A number werebuilt by King Henry II. Henry set about‘improving and repairing’ his French cas-tles in 1161 (Howlett 1889, 209-10), whenthe donjons at Château-sur-Epte andNeaufles-Saint-Martin (both Eure, nearGisors in Normandy), may have beencommenced (Corvisier 1998(2), 147,491); the Norman exchequer accountssuggest that both donjons were completeby 1180 (Stapleton 1840, 72, 110).3 Bothare perfectly cylindrical, without any pro-jections (Fig. 2), and like most Angevindonjons, are unvaulted; Château-sur-Epteshows external offsets which, as in thedonjon at Pembroke, do not always corre-spond to the floor levels.4

Henry II’s cylindrical donjons at Neuf-marché, Seine-Maritime in Normandy(Baudry 2002, 57; Corvisier 1998(2),492-5), and Bourg-le-Roi, Sarthe inMaine (Corvisier 1998(2), 64-71), mayalso have been begun in the 1160s.5 Afifth, at Châtillon-sur-Indre (Indre), inTouraine, is possibly from the 1180s(Corvisier 1998(2), 201; Deyres 1984,364, 374-6); the rest cannot be closelydated (see Table 1). King Richard I continued the tradition atthe important ducal castle of Bonneville-sur-Touques, Calvados in Normandy(Boüard 1966, 355; Mesqui 2013, 24 fig.,167, 215-6; Fig. 3).




Fig. 1 – Northern France in c. 1203 showing circular great towers.


Fig. 2: Two cylindrical donjons of King Henry II: (L-R) Château-sur-Epte and Neaufles (bothEure, Normandy).

Fig. 3: Plan of Bonneville-sur-Touques, Calvados (from Mesqui 2013). Cylindricaldonjon, 1190s (Richard I); towers and probably curtain, 1199-1204 (King John).



Now truncated and used as a private dwell-ing, the Bonneville donjon was under con-struction by 1195.6 It was again unvaulted(Boüard 1966, 355) and, like Pembroke, itwas originally free-standing, on levelground, towards the edge of a pre-existingringwork. The towered enceinte is an addi-tion of King John’s reign (Boüard 1966,371; Mesqui 1997, 456; see below),7 butt-ing obliquely against the base of the donjonin a manner not unlike the secondary cur-tain at Pembroke. Richard also built the cylindrical donjonat Longchamps (Eure, Normandy), where£480 was spent in 1198 (Corvisier 1998(2),393-6; Stapleton 1844, 300, 307, 310,350). He is also thought to have built theround tower on the motte at Le Muret,Cléry (also Eure), which accounted for£420 in 1198 (Pitte 2002, 168 and n. 19;Stapleton 1844, 310), while the similartower at Bonsmoulins (Orne, Normandy)may represent the £419 spent there in1194-5 (Stapleton 1840, 245-6).8 It is sug-gested here that the king’s extensive cam-paign at Pont-de-l’Arche (Eure) included afurther free-standing cylindrical donjon(see below and Fig. 18).9 In Poitou, thefree-standing round tower at Montreuil-Bonnin (Vienne) is also thought to be Rich-ard’s work (Baudry 2002, 57-9 Fig. 4).10

Cylindrical donjons were similarly builtby influential vassals of the Angevin kings,sometimes under their advocacy like thoseat Conches-en-Ouche (Eure), probablybuilt 1180-1200 (Corvisier 1998(2), 255;Landon 1935, 96; Mesqui 1997, 457; Pitte2002, 166), and Brosse (Indre in Poitou),from 1194-1240 (Baudry 2002, 50 and fig.,52, 57-9, 66-7; Fig. 14).11

The donjons of Blois belong to a distincttradition, sometimes featuring vaulted interi-ors and well-represented by the magnificent

tower at Châteaudun (Eure-et-Loir), built inthe 1170s-80s (Erlande-Brandenburg 1970,142). The French Crown built a small num-ber of circular donjons of a rather differentnature (see below), while the form had ap-peared further afield by the 1190s, in easternFrance and the Holy Roman Empire (Mes-qui 2013, 185).Donjons and tours-mixtesThe free-standing cylindrical donjon, then,was firmly rooted in the works of the An-gevin kings. But it was late to arrive in theBritish Isles and, when it did, does not ap-pear to have been employed by the Crown,under whom engaged, peripheral towerswere favoured.12 There is a fundamentaldistinction between these towers and thefree-standing donjon: only in the latter,which was neither defensive (thus normallyunlooped) nor, in many cases, residential,was it possible for access to the space withinto be properly regulated (see Dixon 2002,9-11; Marshall 2002b, 28; Mesqui 1993,252) – all would see it, but few would everenter (Marshall 2002a, 142). The engaged tower, on the other hand,might be used as a residential chamber-tower but could also be equipped for de-fence (the ‘tour-mixte’ of French authors,see e.g. Mesqui 2013, 152-75), becoming amore public space and allowing for multi-ple functions.13 For instance, the cylindricalTour du Moulin at Angevin Chinon (Indre-et-Loire), now thought to have been addedby King John 1200-04, though possibly alittle earlier (Dufaÿ 2011, 88-91, 97-100;Mesqui 2013, 167, 216), was a tour-mixteof this kind. It is engaged with the curtainas a flanking corner-tower, rather than free-standing, and is pierced for archery, but thesix large windows in the upper floor alsosuggest a domestic function, possibly as a‘prospect chamber’.



Fig. 4: The cylindrical donjon at Montreuil-Bonnin (Vienne), thought to have been built byRichard I.

Fig. 5: Plan of Château Gaillard, Eure (from Brown 1976).



King John also built towers of this kind inthe British Isles. The large round tower atDublin Castle (the ‘Record Tower’), is sim-ilarly engaged as a corner-tower which, asat Chinon, appears to have been a tour-mixte with arrowloops (now altered; Man-ning 2003, 72, 84-7; Fig. 8). Also in Ire-land, the circular ‘Reginald’s Tower’ atWaterford, from c. 1207-15, is similarlythought to have been built to flank KingJohn’s castle there, becoming part of thetown defences on the demise of the castlelater in the thirteenth century (Murtagh2017, 163-5). The term ‘dominant tower’may be preferable for these structures, towhich access appears to have been lessrestricted than it was to the donjon.14 Dom-inant towers of this form, with both defen-sive and domestic functions, were adoptedat a number of baronial castles in the BritishIsles during the thirteenth century. Amongthe earliest may be the flanking round tow-er at Barnard Castle (Co. Durham), whichis like Chinon in many respects (Marshall2002b, 27), but articulates with the cham-ber-block to which it is attached, while alsofeaturing a battery of arrowloops.15 Thesemay be contrasted with the narrow open-ings in the body of Pembroke’s donjon,discussed in Appendix 1, which are lightsrather than arrowloops. Though engaged and peripheral, it is likelythat the large semi-round tower built at Rich-ard I’s Château Gaillard (Eure), in 1196-8,was not intended for defence (Fig. 5). Itoverlooks a steep cliff rather than any practi-cal line of assault while, unlike the muraltowers, the body of the tower is not piercedfor archery; functionally, it can be regardedas a donjon (see Marshall 2002b, 32). Fac-ing the interior, the prominent beak showsbuttressed machicolation like that employedby Richard’s father in the double-donjon atNiort (Deux-Sèvres in Poitou), in around

1170-5 (Baudry 2002, 60-1; Mesqui 1993,330); its overall visual impression mayreference the rectangular donjons of hisancestors, as Henry II’s Dover is alsothought to have done (see Brindle 2015, 12;Hulme 2014, 206).16 Jean Mesqui felt thatthe machicolation was yet another aspect ofthe visual ostentation that was the king’sparamount concern at Château Gaillard(Mesqui 1993, 334). As at Chinon, thetower is well-furnished with windowsyielding extensive views. The octagonal tower at Gisors (Eure) is,next to Fréteval, the earliest non-rectilin-ear donjon that can be dated with reason-able confidence, belonging to the reign ofHenry I c.1109-35, and perhaps begun in1123 (Corvisier 1998(2), 328; Howlett1889, 196). The polygonal form was notwidely adopted in twelfth-century France(Mesqui 2013, 187; see Table 1), butfound favour under the Angevins in theBritish Isles e.g. Henry II’s Chilham, Or-ford and Tickhill, and King John’s Odi-ham and Athlone (Goodall 2011, 127-30,167; Hislop 2016, 117).17 The donjon builtin the 1170s-80s at Conisbrough (Yorks.),by Henry II’s half-brother Hamelin ofAnjou, is fundamentally cylindrical and,as built, was free-standing (see Brindleand Sadraei 2015, 6, 29, and plan);18 it hasa similar relationship with the enclosureperiphery, and secondary curtain, to thatseen at Bonneville and Pembroke. Itsshape – a cylinder with six, solid polygo-nal buttresses – was also employed byHamelin in Normandy, in the free-stand-ing donjon at his castle in Mortemer,Seine-Maritime (Brindle and Sadraei2015, 11; Mesqui 1997, 458; Renn 1961,132); their relative dating is uncertain(Goodall 2011, 151; Guy 2013a, 203).19

Whilst it will be shown below that Mar-shal’s donjon at Pembroke has its models



in Angevin Normandy, it also has affinitieswith Conisbrough. Otherwise it has noclose parallels, either in form or detail, inany contemporary castle-work in the Brit-ish Isles where it appears to be the earliesttower of its kind. A rather different trend is apparent in theÎle-de-France during the twelfth centurywhere, while exciting variations in formwere being explored, the plain cylinder wasnot greatly favoured by the French Crownuntil the late 1180s (Mesqui 2013, 185;Table 1). A polygonal shape was chosen atMagny-les-Hameaux, while circular Châ-teaufort, like the similar baronial tower atMaurepas, shows large square turrets; allmay belong to the 1130s-50s (all Yvelines;Corviser 1998(2), 165, 402, 418). The don-jon at royal Compiègne (Oise) is an unbro-ken, unvaulted cylinder (Fig. 6), for whicha similar date has been argued (Corvisier2002, 44), though it could be 1150s-70s(Mesqui 2013, 185).20

King Philip II introduced the cylindricaltour-mixte into the Capetian landscape.The first of these ‘Tours Philippiennes’appears to have been at Bourges (Cher, inBerry), erected in the late 1180s but nowgone (Mesqui 2011, 311 and n. 54; Mesqui2018, 139), closely followed by the Lou-vre, Paris, which can be dated 1190-1202(Fleury and Kruta 2000, 2, 50-1). Thesetowers were isolated;21 subsequent ToursPhilippiennes, beginning in the mid-late1190s at Gisors and Vernon, Eure (Mesqui2011, 311), were engaged with the curtain,furnished with arrowloops and normallyoccupied corners, surrounded by a ditch(Mesqui 2018, 129-39). They were of astark, functional design not generally seenin either the donjons or tours-mixtes of theBritish Isles; large window openings, forinstance, were absent while their rib-vaults,

at several levels, have no contemporaryparallels in Britain (Fig. 7). Not apparentlyceremonial, they perhaps served regionalofficials in a residential capacity, whilebeing actively defensive (Mesqui 2013,162-8).22 In this they differed fundamen-tally from the cylindrical donjon whiletheir double entries, opening to the interiorand exterior of the castle, are also un-known in Britain.The Tour Philippienne was widely used in

the Capetian royal domain after 1202(Mesqui 2013, 162-8; also see Knight 2008,65). Doubtless intentionally, it was symbolicof King Philip’s dominion and four morewere built in Normandy after its capture inmid-1204 (Falaise, Lillebonne, Rouen andVerneuil-sur-Avre) – that is, after the Pem-broke donjon had been commenced. Con-struction continued into the 1220s withDourdan (Essone), the last of Philip’s reign(Mesqui 1997, 153; Fig. 15). And followingroyal precedent, variants of the form becameadopted in other regions of France duringthe early thirteenth century (Mesqui 2013,168, 185). While the first of Philip’s towersmay represent the earliest loopholed tours-mixtes, most of them post-date the similartower at Chinon and their influence on KingJohn’s designs is uncertain.23

Pembroke was Marshal’s only donjon.His other dominant towers – at Kilkennyin Leinster and Usk, Mon. – are peripheral,engaged and defensive tours-mixtes likethose at Angevin Chinon, Dublin and Wa-terford; they are, however, not isolated byditches like Capetian Tours Philippiennes.Marshal is considered to have begun workat Kilkenny Castle, his seigneurial caput inLeinster, in 1207 (Bradley and Murtagh2017, 222; Tietzsch-Tyler 2017, 185), andit seems to have been fairly well-advancedby 1210 when King John visited (Holden



ABOVE: Fig. 6 – Interior of the cylindrical donjon at Compiègne, Oise.BELOW: Fig. 7 – The Tour Philippienne at Vernon (Eure), built by Philip Augustus 1196-1200 (from Mesqui 2011 and 2018). cf. Fig. 15.



et al. 2004, 193).24 The large, engagedround tower at the southern corner (the‘White Tower’; Fig. 8) appears primarilyadministrative, and ‘essentially a publicspace’ (Tietzsch-Tyler 2017, 186) which,like Dublin’s Record Tower, is pierced forarchery (Murtagh 2017, 129-31, 171 fig).25

Influence from Chinon may be implicit –Marshal was at the French castle through-out August 1201 (Hardy 1835a, 3; Hardy1844b, 19-20) – but the Kilkenny towerwas under construction at more-or-less thesame time as the Record Tower, begunsome time after 1204 (Sweetman 1875,35); their immensely thick walls are uniqueto these two towers and reveal a closerelationship between them, if no clear indi-cation of which was the earliest.26 Thecylindrical ‘Garrison Tower’ at Usk, whichoccupies an angle rather than a corner, maybe somewhat later (see below and Fig. 9);it too is pierced for archery (Knight 2008,61 fig., 63 figs.). Unlike Tours Philippi-ennes, these towers appear to have beenneither self-contained residences nor, nec-essarily, chamber-towers, and none lie inassociation with the domestic buildings.27

Marshal at PembrokeA start-date in spring 1201 is now general-ly agreed for Pembroke’s donjon, follow-ing Marshal’s visit in late autumn 1200,with his wife, on his way to Leinster(Crouch 2016, 10-24, 137; Goodall 2011,162). The lordship of Pembroke had re-cently been released to him by King John,coinciding with his first real opportunity totour his western estates. The erection of thedonjon was almost certainly celebratoryand commemorative: he was at last able tomark his marriage, inheritance and eleva-tion to the earldom of Pembroke – all ofwhich were hard-won – at its titular caput(cf. the celebratory donjons at Hedingham,

Dundrum, Rochester etc.; see Appendix 1).It is therefore likely that preliminarygroundwork, recruitment and other prepa-ration commenced immediately after hisvisit. The project’s importance is confirmedby the appointment to the stewardship ofPembroke, in 1202, of one of Marshal’sleading household knights, Nicholas Ave-nel. And the post was no sinecure – Avenelwas recorded at Pembroke in July 1202(Brewer 1863, 227; Davies 1946, 322,327). The donjon may have been more-or-less complete prior to Marshal’s departurefor Leinster in spring 1207: following Der-ek Renn’s formula, the six seasons of workwould account for 18-22 metres (60-72feet) of its total height of 24 metres (78½feet; see Renn 1973, 25-6), although theremay be some evidence of a change of planat parapet level (see Appendix I), possiblyafter Marshal’s return in 1213. The inner curtain overlies the base of thedonjon, at least in its restored form which,judging from the surviving stump againstthe Dungeon Tower, follows the originalline (Figs. 16 and 35). The curtain wastherefore secondary but, like the screen-wall closing off the Wogan cavern beneaththe inner ward, was probably commencedbefore 1207 (see below; Day and Ludlow2016, 66). All evidence suggests that, un-der the Marshals, Pembroke Castle waslimited to the present inner ward, the outerward being an entirely new addition of themid-thirteenth century (as argued in Dayand Ludlow 2016, 67-8, 88-9, 93-4).Marshal in FranceWe will see that in matters of detail, aswell as overall form, Marshal work fol-lows Angevin precedent. Nevertheless, in-fluence from Philippienne designs isgenerally claimed. But the relative chro-nology poses insurmountable problems:



Fig. 8 ABOVE: plan of King John’s Dublin Castle in its final phase (from Manning 2003); BELOW:Ben Murtagh’s reconstructed plan of Marshal’s Kilkenny Castle (from Murtagh 2017).




ABOVE: Fig. 9 – Plan of Usk Castle, Mon. (from Knight 1977).

BELOW: Fig. 10 – Plans of the Angevin Tour du Diable at Gisors, Eure (from Mesqui 2013).


the only Capetian castles that can possiblyhave influenced Marshal were built beforePhilippienne designs took hold.28 For in-stance, he is last recorded in Paris in 1188 –two years before the Louvre was conceived– when he appears to have been an envoy toKing Philip on behalf of Henry II (Holdenet al. 2002, 423).29

Instead, in close company with the An-gevin kings from 1186 onwards (Crouch2016, 68-9), Marshal will doubtless haveseen a number of Angevin cylindrical don-jons. L’Histoire de Guillaume le Maréchal,the biographical poem composed in hishonour in the mid-1220s, implies that Mar-shal may have been with King Henry IIduring the negotiations at Châtillon-sur-In-dre in October 1188 (Holden et al. 2002,399-400; Holden et al. 2006, 109; also seeVincent 2000, 5-6); he had certainly trav-eled extensively in Touraine during the1180s. Henry II’s donjons at Neaufles andChâteau-sur-Epte were lost to King Philipin 1193 (Powicke 1913, 161-2; Stapleton1840, cxlii; Stapleton 1844, cxxviii), butMarshal may have visited them, along withNeufmarché, between 1186 and 1190 whenhe is recorded in eastern Normandy withkings Henry and Richard (Holden et al.2002, 371-2, 375-95; Landon 1935, 26-31).He was moreover at Fréteval in July 1194(Holden et al. 2004, 29-35), which was thenunder Angevin control. Warfare between the English and Frenchcrowns – In which Marshal was active with-in the Angevin lands – dominated the later1190s, ruling out any peaceful contact withthe French court (or recruitment of Cape-tian masons), but yielding many opportuni-ties to witness the building of Richard I’scircular donjons. Marshal was frequently atChâteau Gaillard from 1196 onwards (Cal.Charter Rolls 1226-57, 185; Hardy 1837,15-16, 22, 28-30, 33, 64, 68-70, 96; Landon

1935, 113-42; Round 1899, 195, 379, 473,498), and was at Bonneville-sur-Touquesin January 1200 (Hardy 1837, 32), at leasta year before work commenced at Pem-broke. England and France were again atwar 1202-4, and tensions had been mount-ing since spring 1201 when King Johnbegan strengthening his Normandy castles(Hardy 1835b, 35). While Marshalachieved a détente with King Philip after1204-5, in which – exceptionally – he waspermitted to keep his Norman possessions,the relationship could not be called closeand, apart from a possible, very brief visitin 1216, Marshal was never to see theselands again (Power 2003, 211-13).30

In addition to his experience of theseAngevin cylindrical donjons, Marshalspent at least one night at Châteaudunwhile campaigning in Blois in 1188,(Holden et al. 2002, 409; Holden et al.2006, 110); if the tower was not completeby this time, it will have been well underway. Influence from Châteaudun, at Pem-broke, has been suggested by a number ofauthors (e.g. Avent 2006, 89; Goodall2011, 162; King 1977, 164). Like the don-jon at Bonneville it occupies level ground,and shows certain other resemblances toPembroke – as well as important differ-ences, discussed below. Marshal was only to visit France threetimes after 1203, primarily to meet PhilipAugustus on King John’s behalf. Hecrossed to Normandy in late April-May1204 (Hardy 1835a, 40); journeying fromRouen, still in English hands, he met Phil-ip firstly at Le Bec-Hellouin, Eure (Hold-en et al. 2004, 145), and then at Lisieux,Calvados (Crouch 2015, 163-4; Stapleton1844, cxxxviii). In April 1205, Marshalmet Philip at Compiègne, and then Anet(Eure-et-Loir) near the Norman/Frenchborder (Crouch 2015, 11, 163-4; Holden et



al. 2006, 148). No Philippienne castle ex-isted at these venues; the earlier donjon atCompiègne has been discussed above. Marshal’s whereabouts during his thirdvisit, in March-April 1216, are unknownand our only clue is from Philip’s itinerary,which locates the king at Compiègne, Pont-de-l’Arche, Melun (Seine-et-Marne) andParis during these months (Samaran andNortier 1979, 15-32). The first two wereperipheral to the royal domain and thus amore likely choice of venue than Paris orMelun; the castle at Pont-de-l’Arche is dis-cussed below. Either way, work on most ofMarshal’s castles in the British Isles was,by 1216, well-advanced.Loops, offsets and battersAlso reflecting Angevin precedent, thoughoften attributed to Philippienne influence,is the Marshals’ use of deep basal batters,external offsets, and narrow, simply-splayed triangular loops without internalniches or casemates. While such loops dopredominate in Philip’s work, we have seenthat this belongs mainly to the period be-tween 1202 and 1223. The vast majority ofPhilippienne loops have high sills in theexternal half, and lintelled embrasures(Mesqui 1993, 263 and fig., 281); the for-mer are absent from Marshal work, whilethe latter contrast with their preference forarches. The arrowloop was a comparativelynew innovation which was still evolvingbetween 1180 and 1210 and, while theniched embrasure may have been preferredby the Angevin kings (Mesqui 1993, 253-4), it was by no means universal, particular-ly in Normandy; as John Gillingham notes,Angevin architecture embraced a variety offorms (Gillingham 2001, 118). Moreover,Marshal himself used niched embrasures atKilkenny Castle, 1207-13 (see Murtagh2017, 160, 173).

The Angevin Tour du Diable at Gisors,from the 1180s-early 1190s, shows someloops with niches; others however havesimple triangular embrasures, with full-centred rear-arches, like those in Marshalwork (Mesqui 1993, 254, 265 fig.; Mesqui2013, 263 fig.; Fig. 10). Loops withoutniches were also used in the apices of thebeaked flanking towers at Loches (Indre-et-Loire), in Touraine, which are now at-tributed to Richard I (Héliot and Deyres1987, 48-53, 75), and in the eastern outergatehouse (Porte Saint-Nicholas) at Fal-aise Castle, Calvados (though heavily re-stored), which recent investigation hasshown to be Angevin work of the latetwelfth century (Fichet de Clairfontaine etal. 2016, 245). Triangular loops are also seen in KingJohn’s southern towers at Arques-la-Ba-taille, Seine-Maritime (Langeuin 2002, 356fig., 362; Fig. 13), almost certainly built1201-03 when over £1500 was spent onworks there (Hardy 1835a, 15; Hardy1835b, 69; Langeuin 2002, 362-5; Packard1927, 67-8), as well as in the circular flank-ing towers at Bonneville-sur-Touqueswhich are thought to have been added byKing John before 1204 (Boüard 1966, 371;Corvisier 1998(2), 62-3; Mesqui 1997, 456;Fig. 3). The narrow slits that now survive atRichard I’s Château Gaillard are mainlylights, but a triangular-splayed opening,flanking the postern at the northeast cornerof the inner ward, may be a loop.31

We have seen that William Marshal hadextensive experience of Château Gaillard,influence from which is strongly apparentin his work (see below). Arques-la-Ba-taille was in his custody during 1202 (Har-dy 1835a, 9-11, 15, 22; Packard 1927,67);32 he also revisited Bonneville twice,with King John, in autumn 1203 (Hardy



1837, 32, 111; Holden et al. 2004, 141).33

Loops without niches also appear in theNorman gatehouses at Domfront (Orne),built in the years either side of 1200 (Mes-qui 1997, 152-3 and fig.), and Bricquebec(Manche), from the end of the twelfth cen-tury (Corvisier 1998(2), 84-5); Marshalwas probably with King Richard at Dom-front at Christmas 1198 (Cal. CharterRolls 1257-1300, 173; Landon 1935, 141;Stapleton 1844, xliv), when work on thegate may have already begun (see Staple-ton 1844, lix).34

Loops of this kind were also employedby the Crown in the British Isles, in thenorthwest gatehouse at Dover Castle,Kent (the ‘Norfolk Towers’), at somepoint between the early 1180s and 1216(Brindle 2015, 23; Chapman 2007, 106;Hulme 2014, 220, 226; Fig. 11), and atLimerick Castle, Ireland, where they areuniversal at all levels in King John’s orig-inal work of c.1211-12 (Tietzsch-Tyler2013, 143 fig., 150 fig.; Hulme 2014, 220;Fig. 12). Triangular loops without nicheswere also employed in an ‘Angevin’ mi-lieu by William, son of Hamelin of Anjou(lord of Conisbrough), in the shell-keeptowers at Lewes Castle (Sussex) shortlyafter 1202 (Goodall 2011, 185; Fig. 11).35

Similar loops appear in other baronialwork of the period, for example in thegatehouse and Carrickfergus Tower, from1199-1214, at Warkworth Castle in Nor-thumberland (Goodall 2015, 13, 26, 35)and in the north gate at Helmsley Castle,Yorks., built by Robert de Roos, custodi-an of royal Bonneville-sur-Touques, in1191-1227 (Kenyon 2017, 32-3 and plan).They persisted into the late Middle Ages inwest Wales, where a distinctive regionaltradition developed in Marshal-dominatedareas (Day and Ludlow 2016, 68; see be-low), and are seen in mid-thirteenth-centu-

ry towers at Pembroke itself, at YoungerMarshal Cilgerran, Pembs. (Wiles 2014,191 figs.), and at Kidwelly Castle, Carms.,in work from the 1270s through to theearly fifteenth century (Kenyon 2007, 32fig., plan). Cruciform loops occur in Younger Mar-shal work at Cilgerran and Chepstow (up-per barbican), in Pembroke’s DungeonTower, and the ‘towered keep’ at Carlowwhich are also probably Younger Marshal(see below). Otherwise, their only Mar-shal use is in the outer gatehouse atChepstow (Avent 2003, 62-3), whichmay similarly be post-1219 (see below).They are also seen in Angevin castles,appearing in Richard I’s Bell Tower at theTower of London (Impey and Parnell2011, 21-3; Knight 2018, 164) and inKing John’s outer gate at Kenilworth.36

However, it is not altogether certain thatthey are not later modifications, of earlierloops, at both sites (Neil Guy, pers.comm.), which raises questions abouttheir dating elsewhere. At any rate, theyfirst appeared in France in the Angevinterritories, around 1200, e.g. in the inter-val towers at Cluis (Indre) in Touraine(Corvisier 1998(2), 233; Mesqui 1997,435), at Brosse (Mesqui 1997, 82; seebelow) and, slightly later, under Angevinsponsorship, at Coudray-Salbart andParthenay (both Deux-Sèvres), in Poitou(Knight 2018, 164; Mesqui 1993, 268,272). Moreover, they never really caughton in the French royal domain (Baudry2002, 64; Mesqui 1993, 291-2 and map). Basal oillets make an early appearance inRichard I’s work at Loches (semi-circu-lar), and in Richard or John’s work atMontrichard (Loir-et-Cher) and Dom-front, where they are triangular (Mesqui1993, 272, 284-5, 299; Mesqui 1997, 152-



Fig. 11 – Loops without niches at (L-R) Angevin Dover Castle, 1180s-1216; and Lewes Castle,Sussex, 1202+ (Dover from Brindle 2015).

Fig. 12 – Plan of Limerick Castle (from Tietzsch-Tyler 2013). Nb. thecurrent rectilinear plan was not achieved until the reign of Henry III.



3), a shape adopted at baronial Cluis andWarkworth (see above) and elsewhere.Semi-circular oillets were also used at Bro-sse (Mesqui 1997, 82). The circular oilletsat the Bell Tower, London, are possibly alate- thirteenth-century modification (seeabove), which may have implications forthose present in Marshal work at Chepstowand Pembroke. Oillets of any kind werenotably absent from King Philip’s castles(Baudry 2002, 64; Mesqui 1993, 284-5 andmap), but were adopted by his subjects,under Angevin influence, during the earlythirteenth century e.g. the chamber-towerat Laval, Mayenne (also triangular), nowdated 1218-30 (Chollet and Gousset 2012,261-2; Mesqui 1993, 285). Basal oillets of rectangular form wereused extensively in Marshal work, elderand younger, but are restricted elsewhere.They occur at Pembroke, notably in thedonjon parapet and the Dungeon Tower(n.b. oillets were unnecessary in the slit-lights piercing the body of the donjon), inChepstow’s Middle Ward and upper barbi-can, in the Garrison Tower at Usk, and atCaerleon Castle (Mon.). The last, whichdemonstrably post-dates Marshal’s acqui-sition of the castle in autumn 1217 (Holdenet al. 2004, 391; Jones 1952, 95-6; Knight2008, 60),37 is the only dateable example,which compromises its usefulness in dat-ing the others. They have all been suggest-ed to be late, from 1213 onwards (Knight2008, 60), but Marshal may have employedthe form for some time before it appearedat Caerleon: a rectangular oillet can be seenlow down in the spiral stair to Pembroke’sWogan cavern which, given the overall‘Transitional’ character of the surroundingwork, similar to that in the donjon, wasprobably under construction before Mar-shal left for Leinster in 1207 (Day andLudlow 2016, 66; see Appendix 1). Archi-

tectural detail suggests a similar date forChepstow’s Middle Ward (see below). Atany rate, the rectangular oillet appears tobe an indigenous British feature, and didnot appear in France until the mid-thir-teenth century (Mesqui 1993, 282-3). Dressed external surrounds, as used byMarshal in loops and slit-lights at Pem-broke, Caerleon, Chepstow, Kilkenny andUsk also appear in Angevin work atKenilworth (Lunn’s Tower), Dover(Avranches Tower and Peverell’s Gate),and at London’s Bell Tower where theymay, however, belong to later modification(see above). Plunging embrasures, also em-ployed by the Marshals, were in fairly wideuse by the late twelfth century, in both royaland baronial work e.g. Gisors and Cluis. External offsets, as in Pembroke’s don-jon, Chepstow’s Middle Ward towers andUsk’s Garrison Tower, are seen in An-gevin work both in Normandy and theBritish Isles appearing, inter alia, in HenryII’s donjon at Château-sur-Epte (Fig. 2),his Avranches Tower at Dover and flank-ing towers at Bamburgh, Northumberland(late 1160s-80s), Richard I’s towers atLoches and the Tower of London (1190s),and under King John, in the southwesttower at Arques (1201-4; Fig. 13) andLunn’s Tower at Kenilworth. They also appear in British baronial workof the period e.g. the shell-keep towers atLewes. Absent at Compiègne, they appearin around 50% of Tours Philippiennes(Mesqui 2018) and, while employed atVernon in the 1190s (fig. 7), are mainlypost-1204 as at Chinon, Falaise, Lille-bonne, Rouen and Verneuil-sur-Avre – apattern followed in Philip’s muraltowers.38 Offsets are, however, present insome other cylindrical towers in France,for instance at comital Châteaudun.



ABOVE: Fig. 13 – King John’s southern towers and gatehouse at Arques-la-Bataille (Seine-Mari-time), built 1201-3. Left: plan (from Pitte 2007). Right: the southwest tower, from east.

BELOW: Fig. 14 – Plan of the northwestern mural towers, and donjon, at Brosse, Indre in Poitou(from Baudry 2002).



The pronounced, plain basal batters seen atPembroke and Chepstow are similarly afeature of Angevin design, both in muraltowers and gatehouses – Dover’s inner ward(1180s), King John’s mural towers atArques and Bonneville (Figs. 3 and 13), thePorte Saint-Nicholas at Falaise (Fig. 20),etc. – and in several great towers e.g. Dub-lin, Conches-en-Ouches, and Château Gail-lard where it is very substantial (Fig. 26). Nobatter is present at Compiègne (Fig. 6), norin the Tours Philippiennes at e.g. Chinon,Montlhéry and Vernon; when used inPhilippienne work, moreover, batters areoften shallow as at Dourdan, Falaise, Gisorsand Péronne (see e.g. Mesqui 2018, 132-9;see Fig. 7). A prominent feature at the Lou-vre is the stepped plinth that runs around thebase of all walls and towers (Fleury andKruta 2000, 8, 12, 16, 57); nothing like it ispresent in any surviving Marshal work.39

‘Marshal’ towersA number of well-known and well-definedcharacteristics appear in Marshal work,elder and younger, from which a distinctstyle evolved that was to be highly influen-tial in west Wales for over fifty years (Dayand Ludlow 2016, 68; Goodall 2011, 162,164; also see Wiles 2014); most of them,however, also occur in Angevin work ofthe period. For instance, while the intervaltowers in Chepstow’s Middle Ward showthe ‘standard’ D-shaped plan, the Marshalsalso employed circular towers – not only atcorners, but also in shallower angles as atUsk, Cilgerran, and the Dungeon Tower atPembroke (see Figs. 9 and 16). But in thisthe Marshals were following in royal foot-steps. As noted by Richard Hulme, thesetowers closely resemble those at KingRichard’s Château Gaillard, particularly inits middle ward where they recurve strong-ly to join the curtain, have circular interiors

like the Marshal towers, and none occupiesa corner (Hulme 2014, 217; Fig. 5); KingJohn’s mural towers at Arques are closelycomparable, but show barrel-vaulted interi-ors (secondary?; Langeuin 2002, 356 fig.,361-5; Fig. 13). Similar-shaped towerswere built in similar locations at, for exam-ple, the north curtain at Chinon, by RichardI or John (Mesqui 2013, 77 fig.), at KingJohn’s Bonneville (south tower, see Fig. 3),and later in his castle at Limerick, in thenortheast tower which occupies a length ofcurtain wall between two angles (Fig. 12).Comparable interval towers were built onthe northwest curtain at Brosse by the vis-comte of Brosse, under Angevin authority,between c.1204 and 1242 (Baudry 2002, 50and fig., 52; Fig. 14).40

In Philippienne castles, a preference forD-shaped interval towers is apparent, be-ginning at the Louvre and continuing atFalaise, Péronne, Dourdan and many oth-ers (see Fig. 15), while gate-towers toowere normally D-shaped – a pattern thatwas largely followed in baronial Franceduring the early-mid thirteenth-centurye.g. at Diant, Seine-et-Marne and Mez-la-Maréchal, Loiret (Mesqui 2013, 60-3 andfigs.; Fig. 15). Most Philippienne intervalor gate towers that were circular in plan arelater than 1204, including those at Rouen,at Montargis (Loiret, Île-de-France) from1200-23, and in the contemporary outergate at Yèvre-le-Châtel, also Loiret (Cor-visier 1998(2), 435; Mesqui 1997, 418-19;Mesqui 2013, 22 fig., 43 fig.).41

The pronounced ‘bulging’ outline (orentasis) that characterises the MiddleWard towers at Chepstow, the DungeonTower at Pembroke and the towers atCilgerran can also be seen in Angevinwork, in the southwest tower at Arques(Figs. 13 and 33); it too was adopted



Fig. 15 – ‘Philippienne’ plans. Left: royal Dourdan (Essone). Right: baronial Diant (Seine-et-Marne). Both first quarter of the thirteenth century (from Mesqui 2013).Fig. 16 – Overall plan of Pembroke Castle (from Day and Ludlow 2016).



elsewhere in west Wales, for instance in theinner ward towers, from the 1270s, atKidwelly. None of which however is todeny that certain idiosyncrasies can be seenin Marshal work, elder and younger; noneof them, however, appear in King Philip’scastles, instead contrasting markedly withhis preferred regularity. They include theasymmetry of the tower footprints at Chep-stow, Cilgerran and Usk, with their eccen-tric interiors, and the unequally-sizedtowers in the gatehouses at Chepstow, Car-marthen, and probably Kilkenny (see be-low). Another Marshal attribute and, perhaps,their main departure from Angevin prac-tice, is the rounded rear face seen in someof their towers e.g. the outer gatehouse atChepstow, and in the interval towers at Uskand Cilgerran, which may all be later-peri-od Marshal work. But although the roundedrear face is a defining feature of the en-ceinte at Angers, built by Louis IX duringthe 1230s, and was later employed by theFrench Crown at Carcassonne and else-where (Mesqui 2013, 74-5 and fig.; 354fig.), it does not seem to have been repre-sentative of Philippienne design. It wasagain to prove persistent in west Wales,where mid-thirteenth-century towers atPembroke Castle itself (Fig. 16), at Llanst-effan and Laugharne (Carms.), andLlawhaden (Pembs.), all but stand astridethe curtain.42

Contemporary with the elder Marshal’stowers at Pembroke and Usk are boldlyprojecting, rectangular latrine turrets (Figs.9 and 16). Similar rectangular latrine tur-rets also appear in Angevin work of c.1180-1204 at Château Gaillard, Chinon and Fal-aise (Mesqui 1993, 176; Mesqui 2013, 39fig., 123 and fig., 264 fig., 296), but werenot a feature of Philippienne design (Mes-

qui 1993, 179-80). Like other featuresmentioned above, they became widelyadopted in thirteenth-century west Wales,from which they spread to other regions ofWales, and beyond (Ludlow 2018, 268-9).Geometric planningKing Philip II’s castles are noted for theirgeometric planning, which was subse-quently adopted by his successors at e.g.Saumur, Maine-et-Loire, built 1220-30(Mesqui 1997, 348), and by his barons ate.g. Diant and Mez-le-Maréchal, Loiret(Mesqui 2013, 47 and fig., 60-1 and fig.,63; Fig. 15). It has also been suggested atMarshal’s Kilkenny and Usk (Knight2018, 164; O’Keeffe 2009, 287 and n. 37;Sweetman 1999, 50). The earliest geometric castle within theCapetian and Angevin ambit is thoughtto be Druyes-les-Belles-Fontaines(Yonne), an important pre-Philippiennecastle dated to the 1170s (Fig. 17); asthe residence of Peter de Courtenay,Count of Nevers and a cousin of theFrench king, it undoubtedly influencedPhilippienne planning (Mesqui 2013,266-7). Quadrangular and symmetrical,with circular corner towers, it showsneither donjon nor dominant tower(Mesqui 2013, 40-1 and fig., 60). Mar-shal passed close by, or possibly eventhrough Druyes with kings Richard andPhilip, on their way from Donzy(Nièvre) to Vézelay Abbey (Yonne) forthe crusader conference in July 1190(Landon 1935, 36; Round 1899, 35);Peter de Courtenay accompanied thekings to Palestine, and the right of hos-pitality at Vézelay, claimed by theCounts of Nevers (Huygens 1976, xxvi,38, 47, 229), was doubtless significantduring the royal stay at the abbey.43



The two kings arrived in Palestine thefollowing summer,44 where quadrangularcastles without donjons had also been built.Philip spent less than fifteen weeks there, allof them at Acre apart from three days atTyre (Runciman 1951, 34, 52). Richard’sexperience, though longer, was confined tothe Kingdom of Jerusalem and specificallythe coastal strip between Acre and Ashkelon(now in Israel). The castles he is known tohave seen, some of which he began rebuild-ing, were concentrated around the latter andwere mostly rectangular enclosures withfour square corner towers, dating from the1130s onwards (Kennedy 1994, 31-2;Landon 1935, 57-62; Pringle 1989, 14, 17-18).45 The circular corner towers at Druyesmay be the earliest in either France or Brit-ain where flanking was, until the 1190s,mainly an Angevin feature, and the curtainsat Henry II’s Loches, Gisors, Rouen and LeChâtelier (Indre-et-Loire, Touraine) appar-ently show the earliest rounded intervaltowers (Corvisier 1998(2), 184; Héliot andDeyres 1987, 24, 42; Mesqui 2013, 37, 43,69, 259-61). All predate the adoption of therounded tower into mainstream Crusaderdesign (see below). At least one castle of Richard I, begun onhis return, appears to show geometric plan-ning and a circular corner tower. Nearly£1000 was spent by the king at Pont-de-l’Arche in 1195 (Stapleton 1840, 137, 156,236-7).46 The castle, which fell to Augustusin mid-1204, has now gone but the sourcesshow that it was not of standard Philippi-enne design. It comprised two baileys in anarrangement not unlike that at ChâteauGaillard (Fig. 18); the inner ward, howev-er, was a regular rectangle. The question ofhow much of this work might have beenRichard’s, and how much Philip’s, is still amatter of debate (Launay 2015), includingthe freestanding donjon (see above); the

spatial relationship between the latter andthe curtain wall is however paralleled atBonneville, and later at Pembroke (seeFigs. 3 and 16). Also suggestive of Rich-ard’s hand are the solid, semicircular gate-towers, similar to those in the Porte desChamps at Falaise, built by the Angevinkings in the later twelfth century (see be-low), and a circular interval tower likethose discussed above.47 Richard visitedPont-de-l’Arche twice in summer 1195(Landon 1935, 102, 104), perhaps an indi-cation that work was underway.The Norman exchequer rolls, combined

with recent research – much of it French –show Richard I to be one of the busiercastle-builders of the medieval world (if notso in Britain): in addition to the above-men-tioned sites he spent heavily in France, forexample at the eastern Normandy castles ofBellencombre, Boutavant, Courteilles,Gamaches, Lyons-la-Forêt, Moulineaux,Orival, Radepont, Tillières-sur-Avre, Vau-dreuil and Verneuil-sur-Avre (Pitte 2002,166; Powicke 1913, 281-2; Stapleton 1840and 1844, passim).48 But much of this workhas been lost or altered beyond recognition;we have only a partial glimpse of the fullrepertoire of his designs. It is also still uncertain how much Rich-ard, and western Europe generally, mayhave owed to experience of crusader Pal-estine (see inter alia Goodall 2011, 162;Hulme 2014; Kennedy 1994, 186-9; Mes-qui 2013, 37-8, 41, 264-6; Pringle 1988,xxi-xl; Pringle 2014, 48-56). For instance,while the main body of work at the Louvrehad to wait for King Philip’s return fromcrusade, at the end of 1191, to get fullyunder way (Mesqui 2013, 162), it had beencommissioned before he departed (Fleuryand Kruta 2000, 2), including, presum-ably, its geometric layout – which, as



Fig. 17 – Axonometric view of the castle at Druyes-les-Belles-Fontaines(Yonne), built in the 1170s; the gate-tower is later (from Mesqui 2013).

Fig. 18 – Plan of the castle at Pont-de-l’Arche, Eure (from Mesqui 2013)



Fig. 19 – Coudray-Salbart, Deux-Sèvres. Left: in c.1202-6. Right: in 1242 (adapted from Mesqui2013).

Fig. 20 – L-R: The Porte Saint-Nicholas at Falaise, built by Henry II or Richard I, and Chepstow’sMiddle Gate.



noted above, was probably derived directlyfrom Druyes. Marshal too had spent time inthe Middle East, having undertaken a pil-grimage between early 1184 and early 1186(Crouch 2015, 9; Holden et al. 2006, 105-6). But, while this potentially means expe-rience of eastern fortification, our onlysource for the visit is the biographical His-toire, which is silent about Marshal’smovements to, from and within Palestine(Holden et al. 2002, 371).49

King John had visited Paris in the early1190s and 1201 but, while geometric plan-ning is evident in the later Angevin castlesat Dublin and Coudray-Salbart, it may notresult directly from Philippienne influence:Richard I may have provided models closerto home (see above). The Tower of Lon-don, too, was by 1200 a markedly rectilin-ear castle, even if not ‘correctly’ geometric(Impey and Parnell 2011, 22 fig.).50

Dublin and Coudray-Salbart were, more-over, the result of gradual development. Inits final phases, before it was lost to theFrench in 1242, Coudray-Salbart was arectangular castle of great regularity. Workprobably began in 1202, paid for by subsi-dies from King John (Curnow 1980, 44-5).John was in Poitou from June until October1206, staying at Coudray itself on 22 Sep-tember, and at nearby Niort in August andOctober (Hardy 1833, 73-4; Hardy 1835a,667) so further visits are possible. Howev-er, Coudray did not become geometric untilit was enlarged in Phases 3-4, which maybelong to the second or third decades of thethirteenth century; the castle John saw in1206 was much smaller and of irregularplan (Mesqui 2003, 79-80; Fig. 19). It ishowever possible that Phase 3 work mayhave started, at least, when John was inPoitou during his 1214 campaign and,though he is not recorded at Coudray itself,a visit cannot be ruled out.51

The transformation of Coudray into ageometric castle may have coincidedwith the establishment of a rectilinearplan at King John’s Dublin. We sawabove that there is some uncertaintyregarding the constructional history ofthe castle but, as at Coudray, develop-ment was incremental; it only assumedits present form after 1213, with workcontinuing until around 1230 (Manning2003, 72-3; Fig. 8; Tietzsch-Tyler 2018,121). Staged development has also beenrecognised at John’s Limerick Castle,which did not achieve its current geo-metric plan until the reign of Henry III(Tietzsch-Tyler 2013; Fig. 12). Ben Murtagh has pointed out that anyrectilinearity at Marshal’s Kilkenny –which, at present, is uncertain – is proba-bly incidental and dictated by the underly-ing, earlier ringwork (Murtagh 2017, 173;see Fig. 8). The same appears to be thecase at Usk which is not truly rectangular,showing a sub-oval plan defined by tenlengths of straight walling, similarly fol-lowing a preexisting earthwork enclosure(Fig. 9);52 asymmetrical in outline, theenceinte incorporates an earlier rectangu-lar tower (Knight 2008, 57-8). Marshal’s work at Usk is difficult to date;a date after 1213 has been argued on thebasis of the arrowloops (Knight 2008, 60),the uncertainties over which have howev-er been discussed above. Nevertheless theearl stayed at Usk in July 1217, and at hiscastles at Chepstow and Goodrich, Herefs.(Hardy 1833, 314; Pat. Rolls 1216-25,79), and while this visit was primarily aresponse to the growing threat from theneighbouring Welsh lord Morgan ap Hy-wel of Caerleon (Holden et al. 2006, 186-7), he was again at Usk in December 1217(Hardy 1833, 348). The visits may coin-cide with building work in at least one of



these castles – namely Goodrich, whichpossibly does show evidence of geometricplanning. It has long been thought that the founda-tions of an early, circular tower at thesouthwest corner of Goodrich, and thelower courses of its east curtain, belongto the Marshal period (Ashbee 2009, 30;Radford 1958, 5, 8). According the His-toire, the castle was attacked in October1216 (Holden et al. 2004, 269; Holden etal. 2006, 166), which may have triggeredbuilding work; the earl is again recordedat Goodrich in September 1218 (Hardy1833, 370), perhaps providing an oppor-tunity to check its progress, which mayhave continued under his sons. The foun-dations show a plain, circular corner tow-er which is otherwise little-known(Shoesmith 2014, 175); its relationshipwith the east curtain however suggests acompact, geometrical castle defined bythe present square ditch. Nevertheless,Ron Shoesmith considers that no workwas undertaken at Goodrich under theMarshals, and that the castle retained itsearlier, uncertain form until the late thir-teenth century, when he suggests thepresent ditch was cut and the corner tow-er was built (Shoesmith 2014, 22, 106-8,117-18, 129-30).53

If Goodrich does feature Marshal-periodgeometric planning, the long passage oftime since the earl’s experience of eitherPalestine or Druyes-les-Belles-Fontainesmust be considered. Neither is then likelyto have been a direct influence which,instead, was again probably derived fromthe Angevin Crown.54

Gates and gate-towersMarshal castles show a variety of arrange-ments at the main entry but, by and large,they reveal a preference for relatively sim-ple forms. Chepstow’s Middle Gate is astraightforward doorway with one pair ofgates and no portcullis (Avent and Turner2006, 64-5). Its detail, which survives, isin the ‘Franco-Angevin’, TransitionalGothic style employed throughout north-ern and central France during the latetwelfth century (Fig. 20). It can be com-pared with that in the Porte Saint-Nicho-las at Falaise, built by Henry II or RichardI (see above) – particularly in its cham-fered jambs and imposts – and in the entryat Bonneville-sur-Touques. The Chep-stow jambs show pyramidal chamfer-stops in a very early context, but barredstops were used in King John’s work,from 1201-4, in the Gloriette at CorfeCastle (Dorset), where the triangular-headed rear-arches were emulated byMarshal in Chepstow’s upper bailey tower(see Turner 2006, 79-80). So it may bethat the Middle Gate was begun soon afterMarshal’s return from Normandy in late1203. It has been suggested that the outer wardand its curtain were already in existencewhen the Middle Ward was built, becauseof the disposition of loops in its southtower (Avent and Turner 2006, 64; therectangular oillet here is discussed above),but this may be incidental; the fact that theMiddle Ward east curtain was flanked bytwo towers, the southern of which is acorner tower, suggests that it was an exter-nal line of defence when built. Straightcurtains were not unknown by c.1200 (cf.London, Warkworth etc.), while Chep-stow Castle occupies a ridge across whicha straight defensive line is easiest.



The gate at Usk is also a simple doorwaybut here the arch is notably plainer, with asingle order, while moulded detail is ab-sent. And, unlike Chepstow, there is a port-cullis (Knight 2008, 65), possibly operatedfrom the parapet as in the Younger Marshalpostern at Cilgerran (see Hilling 2000, 21).The main gate at Cilgerran, probably fromthe 1230s, is a simple, square tower,though showing two portcullises (Hilling2000, 13-14, 18). Little remains of thesimilar gate-tower at the Marshals’ TenbyCastle, Pembs., which is difficult to dateclosely. It is not unlike Cilgerran’s in gen-eral form, but is considerably smaller andshows arrangements for just one portcullis.It was later given a barbican enclosure. The Phase 1 gatehouse at Coudray-Salbart(c. 1202-06) was a semicircular tower with astraight passage (Fig. 19), a plan that was notwidely adopted in France and, where occur-ring, is mainly late thirteenth-century or lateras at Bressuire and Thouars, both Deux-Sèvres (Mesqui 2013, 314-5). However, itwas employed, probably by William Mar-shal II, in the inner barbican of the Marshalcastle at Dunamase (Co. Laois) in Ireland(Hodkinson 2003, 33 fig., 41-2, 48).55

A gatehouse of similar plan was built, ataround the same time, at Hubert deBurgh’s Grosmont Castle (Mon.), althoughhere it extends into the bailey (Knight2009, 9 fig., 20-1). It is likely that, duringhis long periods in Poitou 1202-4, and1213-1214/15 when he was seneschal ofthe province, de Burgh saw the Phase 1Coudray gate-tower before it became sub-sumed within the later work there. So hemay have been the source for Dunamase,as a close ally of William Marshal II until1226, when the latter supported a rival’sclaim in Ireland costing him the justi-ciarship there (Crouch 2015, 20).

Angevin design is explicit in the southeastgate built in 1201-4 at Arques, which wasa single tower, of circular plan, with astraight passage. Now destroyed, it project-ed wholly from the curtain into which, likethe mural towers, it strongly recurved(Langeuin 2002, 361 fig., 363; Fig. 13).56

A similar gate-tower was built, before1204, in Falaise’s inner ward (Fichet deClairfontaine et al. 2016, 239, 246) andlater, during Phase 4, at Coudray-Salbart(Fig. 19). St John’s Tower at Dover fol-lows the same pattern, though isolated on aforework, and may be roughly contempo-rary with the Coudray gate-tower havingbeen added between 1217 and 1221 (Brin-dle 2015, 23, 44 fig.; Fig. 11). A cylindri-cal gate-tower was also employed byWalter de Lacy at Trim Castle, Co. Meath,in 1201-07 or 1220-23 (McNeill 1990, 329fig., 331, 334; O’Keeffe 2017, 44-7); un-like Arques and Coudray it stands astridethe curtain, as it may have done at Falaisewhere Walter was recorded in June 1200(Hardy 1837, 69; see below).57 The formwas later adopted in other regions of north-ern and western France, persisting into thefifteenth century (Baudry 2002, 53; Mes-qui 2013, 312-16). A striking feature of Marshal Pembrokeis the large, D-shaped gate-tower or‘Horseshoe Gate’ (Fig. 16). Like the innercurtain to which it is attached, it is sec-ondary to the donjon and was probablybegun 1204-7 (Day and Ludlow 2016,66); now truncated, it lacks evidence for aportcullis.58 It was a hollow tower, pro-jecting wholly from the curtain and en-tered through its flank, with a second entryat right-angles. Similar forms occurredthroughout the Mediterranean and NearEast, and the design was adopted in thetwelfth century by the Crusaders (Kennedy1994, 59-61, 91, 150-2; Pringle 1988, xxx,



xxxiv, 32-3; Pringle 1989, 223); it has beensuggested that it provided a model for thePembroke gate, via Marshal’s pilgrimage inthe early 1180s (Avent 2006, 89-90; King1977, 164-5; King 1978, 107 and n. 90; Prin-gle 1988, xxxviii n. 56; Pringle 2014b, 48, etal.). But all these towers are square in plan.Significant questions arise from the twenty-year gap before the design was employed atPembroke, its use in a rounded tower and,perhaps most crucially, the presence of a verysimilar D-shaped gate-tower at the castle ofSaranta Kolones in Paphos, Cyprus. Saranta Kolones was begun after the con-quest of Cyprus by Richard I, in May 1191,on his way to Palestine (Pringle 2014b, 48).Much of the work appears to have beenundertaken by the Lusignans, kings of Jeru-salem and Cyprus, between 1192 and 1205(Avent 2006, 90 and fig; also see Runciman1951, 44-7, 84, 103), but was the castleplanned in summer 1191, under Richard I’sadministration? Its concentric plan pre-exist-ed at Henry II’s Dover (as noted in Pringle2014b, 48). The rounded tower moreoverwas almost unknown in existing Crusaderwork (see e.g. Kennedy 1994, 18, 85-96,115; Pringle 1988, xxxii-xxxiii) and, whileArmenian influence may lie behind its use atPaphos (Pringle 2014a, 368-70),59 both gatesare, in form, very like the D-shaped muraltowers at Angevin Gisors (cf. Figs. 10 and16). Might it be a western adaptation, underKing Richard, of an eastern design? A connection with Richard I might explainMarshal’s adoption of the design.60 Andwhile Paphos may more likely be post-1192, Richard may yet have provided theconduit to Pembroke: Geoffrey de Lusig-nan returned from Palestine and was inter-mittently in the king’s retinue, withMarshal, 1196-99 (Landon 1935, 113, 123,144; Runciman 1951, 84). Nevertheless,the Pembroke gate exhibits another eastern

feature: a socketed pivot-stone, withinwhich the gates were hung (King 1978,106; Fig. 21). Like right-angled entries,pivot-stones had long been widespread inMuslim fortifications in the Mediterraneanand Near East, and were adopted by theCrusaders during the twelfth century (Fig.21).61 While Marshal may have seen simi-lar pivot stones in the early 1180s, the useof such an engineering feature at Pem-broke is likely to represent a mason’s pref-erence rather than a patron’s instruction(Malcolm Hislop, pers. comm.). Richard Iundertook building works at a number ofcastles in the Kingdom of Jerusalem (seeabove), and his masons will doubtlesshave been familiar with the feature, whilethe presence of masons with experienceunder the Angevin Crown, at Marshal’sPembroke, is argued below.62

David King considered it possible thatanother ‘horseshoe’ gatehouse was builtby the Marshals, after 1217, at CaerleonCastle, on the strength of two drawingsfrom c.1800 (King 1977, 165). MalcolmHislop has recognised that a fourth exam-ple exists at Montgomery Castle, built byHubert de Burgh 1223-8 as a postern to hisinner ward (Hislop 2016, 168-9; Fig. 22).While independent Angevin influence ispossible, the initially close relationshipbetween de Burgh and the younger Wil-liam Marshal must be taken into account.It may be significant, too, that WilliamMarshal II was responsible for administer-ing scutage in Montgomery in early 1225(Hardy 1844a, 16, 28, 30), while his clerkPeter of Luton disbursed payments forworks there, on behalf of the Crown, in1224-5 (Hardy 1833, 586, 591, 597, 617;Hardy 1844a, 28, 31).63

The adoption of the ‘horseshoe’ form ata fifth castle, Caldicot (Mon.), probablyduring the 1230s (Guy 2016a, 161, 174),



Fig. 21 – Left: the lower pivot-stone in Marshal’s ‘horseshoe’ gate, Pembroke Castle. Right: theupper pivot-stone for the gate at Krak des Chevaliers, Syria (photograph: Jean Mesqui).

Fig. 22 – Plan of the inner ward at Montgomery Castle, 1223-8, showing the ‘horseshoe’ posterngate-tower at the north end (adapted from Knight 1993).



may be attributed to the dominance of theyounger Marshals at nearby Chepstow(King 1978, 106-7).64 Two further exam-ples may exist in the north of England. Thebailey at Sandal Castle, Yorks., is occupiedby a thick-walled, D-shaped structure lead-ing to the donjon, entered from the side andshown as a tower in a sixteenth-centurydrawing (Hislop 2016, 174 and n. 6; Ke-nyon 1990, 81). It was probably builtc.1265-71 by the earl of Surrey, John deWarenne (Hislop 2016, 174), whose moth-er was William Marshal’s eldest daughterMatilda (Crouch 2015, 39);65 the sixtyyears separating it from Pembroke musthowever be taken into account. The gate-tower at Barnard Castle is similar, thoughruinous and yet to be closely dated (Fig.23); dates between the twelfth and fifteenthcenturies have been suggested (see Guy2013b, 46, 56), though construction in thelate thirteenth century and a relationshipwith Sandal is favoured by Malcolm His-lop (pers. comm.).66

Twin-towered gatehousesIn contrast with the above entries, Chep-stow’s outer gatehouse is a complex, twin-towered structure. Its doors, which sur-vived in situ, yielded a felling-date of1159-89; other features of the gates sug-gested that joinery had taken place shortlyafter felling, while the timber was stillgreen, and had not been re-used in the gates(Avent and Miles 2006, 52). So the en-trance arch within which they hung appearsto have been built fairly soon after Marshalgained possession of Chepstow in 1189.But Neil Guy has shown that the gatehouseis the product of incremental development,probably extending into the early four-teenth century (Guy 2016a, 192-5). Whilethe entry itself may be the work of the elderMarshal – along perhaps with the basic

disposition of the towers67 – the chambersover the passage, and the rear section, ap-pear to be later additions. It is even possiblethat the towers themselves were not com-menced until Marshal’s son succeeded in1219 (Guy 2016a, 195 and n. 18); theircruciform loops are normally a feature ofYounger Marshal work (see above), whilethe spiral-stair shaft protrudes into theground-floor chamber of the southern tow-er, as in the post-1219 Dungeon Tower atPembroke (see below) and Cilgerran’s westtower. But the dating of the original entryrelative to the Middle Gate – which wassuggested above to have predated the outerward – remains problematical. Another twin-tower gatehouse was for-merly to be seen at Kilkenny (Fig. 8), butwas demolished in the eighteenth centurymeaning that, although it may have beenbegun by Marshal c.1207,68 we cannot besure whether its constructional history wasequally complex. Ben Murtagh and DanTietzsch-Tyler agree however that thetowers were unequally-sized, as at Chep-stow (Murtagh 2017, 171; Tietzsch-Tyler2018, 126); Murtagh also suggests thatthey had round backs and lacked any ad-joining rear-section, as in the intermediatestages at Chepstow. While Marshal’s oth-er work is notable for its conservatism,rather than innovation, Kilkenny’s impor-tance as his new caput in Leinster must betaken into account, just as Chepstow washis favoured castle in England and Wales(see Appendix 2). But an asymmetric,twin-towered plan, with one tower largerthan the other, is also seen in the gate-house at Carmarthen Castle, which wasalmost certainly built by the younger Wil-liam Marshal, in the mid-1220s, rather thanby his father (Ludlow 2014, 183-4, 189,199-202). The gatehouse was rebuilt in



1409-11, but following the footprint of itspredecessor in which the towers also re-curve strongly into the curtain (but not theentry; Fig. 24). There is moreover little indication in thesources that the elder Marshal showed anygreat interest in his British estates before hewas created earl in 1199 and he preferred,when not on Crown service, to reside inNormandy until its loss in 1204 (Crouch2016, 96-8; Power 2003, 211). And it washere, in his caput at Longueville-sur-Scie(Seine-Maritime), that Marshal himselfmay have built a twin-towered gatehouse.The towers are rounded but are now in verypoor condition, difficult to interpret, andsubject to very little published analysis(Jean Mesqui and Jacques le Maho, pers.comm.; Fig. 25).69 They are solid as atPont-del’Arche, Falaise’s Porte desChamps, and at Hamelin of Anjou’s Conis-brough and Castle Acre (Norfolk), whichmay all be from the 1190s (see above; Guy2016a, 151), but appear to recurve behindthe entrance passage. They also show verydeep offsets, as in Richard I’s Bell Towerat London; it is even possible that, like theBell Tower, their lower stages were polyg-onal (cf. Domfront). So the gatehouse –which lacks a portcullis – may be from1194-1203, when Marshal was primarilyresident in Normandy.70

Twin rounded gatehouse towers neverthe-less had strong Angevin precedents. Inaddition to the solid Porte des Champs, thenear-contemporary Porte Saint-Nicholas atFalaise has hollow, D-shaped towers, ap-parently with closed backs (Fichet de Clair-fontaine et al. 2016, 232, 244-5; Fig. 20).Marshal is known to have been at Falaisein January 1203 (Holden et al. 2004, 137),and may have stayed there with Richard Iin late December 1198 (see Cal. Charter

Rolls 1257-1300, 173; Landon 1935,141). Marshal probably also had experi-ence of the gatehouse at Domfront, asnoted above; probably begun around1198, it has two D-shaped, but seven-sid-ed towers, with closed backs (Mesqui1997, 152-3 and fig.).71 The two lattergatehouses are comparable with the north-west gate at Angevin Dover, built betweenthe early 1180s and 1216 (Brindle 2015,23), where the D-shaped towers appear tohave open backs like the similar gate-house at Pevensey, Sussex, which isthought to have been built by Richard I(Chapman 2007, 102-9); here, however,the towers are elongated in a form thatwas later adopted – with closed backs – atKenilworth (c.1210-15) and Dublin (per-haps 1220s; Fig. 8).72

King John’s twin-towered gatehouse atLimerick Castle can be dated to c.1211-12(Tietzsch Tyler 2013, 145). It is not unlikeChepstow in the sheer scale of the towers(D-shaped here, though with circular inte-riors), and its triangular loops (Fig. 12).Influence in the other direction, from Mar-shal to the king’s work, is however unlike-ly: John’s suspicions of the earl,beginning in 1205, did not fully abateuntil he was recalled from Ireland in1213,73 while we have seen that the Chep-stow gatehouse may be the later of thetwo. Moreover, the Limerick gatehousediffers from Chepstow in its overall ar-rangements, with a contemporary cham-ber over the gate-passage, which extendsinto the bailey. An early context for thesimilar gatehouse at Carrickfergus Castle,Co. Antrim, has recently been questionedand a start date after 1210, under KingJohn, now seems likely – perhaps ac-counting for the £500 spent at the castle1217-22 (Tietzsch-Tyler 2018, 130).



ABOVE: Fig. 23 – The ‘horseshoe’ gate-tower in the inner ward at Barnard Castle, Co. Durham.

BELOW: Fig. 24 – Plan of the gatehouse at Carmarthen Castle (first floor), showing asymmetricaltowers: early fifteenth-century on a footprint from the 1220s (from Ludlow 2014).



Architectural detailMarshal-built carved detail only survives insitu in Chepstow’s Middle Gate and atPembroke, in the Transitional Gothic win-dows of the donjon and Wogan screen-wall. The donjon was commenced in 1201,influenced by Angevin prototypes in north-ern France (see above) including the starksimplicity of its overall treatment. The de-tailing and dome however show specificFranco-Angevin attributes: it is likely that amaster-mason(s) from Normandy, Anjou orTouraine, or with direct experience underthe English Crown there, was personallyresponsible. If not present from the first, hisrecruitment may have occurred after thefinal loss of these territories in 1204-5,when a number of craftsmen who hadserved the Crown in Normandy were re-corded in the British Isles (Hislop 2016,151; Hulme 2014, 225-6). It was observedabove, moreover, that the Pembrokemason(s) may have had direct experienceof royal works. Lindy Grant has shown thatthe Angevin kings used native artisans inNormandy (Grant 1994, 78); at least onewas from the vicinity of Château Gaillard,and was in King John’s service after 1204,the master-carpenter Nicholas de Andeli(Colvin 1963, 61-2; Hislop 2016, 153 n. 15;Hulme 2014, 225).74 In this context, it maybe significant that Marshal was in Pem-brokeshire during late 1204, and DavidCrouch suggests that he spent some weeksin the area (David Crouch, pers. comm.),where he probably issued his grant toMonkton Priory, Pembroke (Caley et al.1846, 320-1; Crouch 2015, 161-2); thecomposition of its witness-list suggestsMarshal at least visited Pembroke Castle toview the ongoing work there.75

There was no motte at Pembroke for thedonjon to occupy and, as at Angevin Bonn-

eville and Neufmarché – and e.g. Fréteval –it lies on level ground. In this it is also similarto the ‘Angevin’ donjon at Hamelin of An-jou’s Conisbrough. Conisbrough is also theonly pre-existing donjon of this kind inwhich a large, twin-light window, directlyover the entry, survives – as at Pembrokewhich it may have influenced (noted byGoodall 2011, 163; Figs. 26 and 27). Bothdonjons also show a large third-storey win-dow; in design, however, the windows arerather different at the two sites. The Pembroke windows appear ultimate-ly to be derived from the Franco-Angevinpattern seen at Château Gaillard (Fig. 26),and in Henry II’s work in, for example, therectangular donjons at Portchester and Ap-pleby, Westmorland (Goodall 2011, 37fig., 137 and fig.). The twin lights at allfour occupy plain, arched surrounds withsolid tympana, and broad mullions –chamfered at Pembroke and Château Gail-lard – take the place of cylindrical shafts;Pembroke and Château Gaillard also shareslightly pointed outer arches. The lightshowever are lintelled in the Angevin win-dows (‘Type B3’, see Mesqui 1993, 234fig.), unlike the voussoired heads em-ployed at Pembroke which are pointed onthe second floor, and semicircular at third-floor level (Mesqui’s Type B2). The over-all disposition of Pembroke’s windows,but not their form, is broadly paralleled inthe tower at Châteaudun (Mesqui’s TypeA2), visited by Marshal in 1188. The figurative carved heads in the Pem-broke tympana are demonstrably derivedfrom those in the windows of the rectangu-lar donjon at Chambois in Normandy (Or-ne), built by William de Mandeville, Earlof Essex, before his death in 1189 (Mesqui1993, 217-8 and fig.; Mesqui 1997, 102;Fig. 27). Marshal purchased Chambois



ABOVE: Fig. 25 – Marshal’s castle at Longueville-sur-Scie (Seine-Maritime), Normandy. Left:sketch plan. The rectangular block behind the gatehouse is secondary (the large southwest flankingtower, at least, is late-medieval). Right: the southern gatehouse tower, from northeast, in 2017.

BELOW: Fig. 26 – The donjons at (L-R) Conisbrough (Yorks.) and Château Gaillard (Eure).



Fig. 27 – Left: the second-floor window in the donjon at Pembroke (photograph: Neil Guy). Right:a window in the donjon at Marshal’s Chambois (Orne). Note the figurative carved heads in thetympana at both.

Fig. 28 – Pembroke Castle from the northwest in c. 1775 (detail from a drawing by RichardWilson; National Library of Wales).



from King John while in France duringMay-October 1200 (Hardy 1835b, 39), butlost it to Philip Augustus in early 1204(Power 2003, 208). Here, the lights alsohave slightly pointed heads, in pointed out-er arches, but are cut from solid lintel-slabs(Mesqui’s type B5). While there are no close parallels for thePembroke window design in the BritishIsles, aspects of its detail suggest someEnglish influence from the West Country,namely the use of dogtooth ornament in thesecond-floor window.76 Dogtooth waswidely used in England and Wales duringthe late twelfth and earlier thirteenth centu-ries, and can be seen nearby in the navetriforium at nearby St David’s Cathedral(Pembs.), from 1182-1200, which belongsto the ‘West Country School’ of architec-ture (defined by Brakspear 1931; Thurlby2006, 307-8, 328-35). It was, however, un-usual in Angevin France (Hourihane 2012,300).77 The absence of capitals in the Pem-broke window-surrounds may also betrayWest Country influence (see Brakspear1931, 6-7), although it can also feature inthe Franco-Angevin tradition (Mesqui1993, 202 fig., 234 fig.). So it is possible that, like the oolite fromwhich its dressings were cut, at least someof the donjon’s free-masons were drawnfrom Marshal’s West Country heartland.78

An eighteenth-century view of the castlesuggests that the main first-floor entry hada semicircular-headed surround, possiblywith continuous roll-moulded order(s) inthe West Country style (Fig. 28; also seeRenn 1968, 38), like those surviving in theentries at nearby Monkton Priory and itsdependent church of St Mary, Pembroke,which are similarly in oolite (Fig. 29).79

However, the depiction may be convention-alised: most surrounds appear to be robbed-

out as today, detail is otherwise very dif-ferent from Monkton and St Mary’s andall surviving surrounds, apart from in thewindows, are very plain. The twin-lightwindow in the Wogan screen-wall is verylike those in the donjon and is clearlyclose in date, ie. 1204-7; it occupies asemicircular outer arch, the lights havetwo-centred heads, but dogtooth andcarved heads are absent.Vaulting and domesAlthough King John used vaulting in theTour du Moulin at Chinon, it was notwidely used by the Angevins or their bar-ons in either donjons or tours-mixtes (Ta-ble 1) and, when occurring, is normallyconfined to basement level e.g. Conis-brough and Barnard Castle.80 It was morecharacteristic of Blois, where it could alsobe used on the entry floor as at Château-dun (Mesqui 2013, 137 fig., 214-15), butsuch expansion of vaulted space remainedunusual in France until adopted by PhilipAugustus in the late 1190s (see below).81

Châteaudun also has a summit-vault,82

though with a flat surface rather thandomed as at Pembroke. Nothing quite likethe Pembroke dome exists in other cylin-drical donjons in either France or the Brit-ish Isles except, perhaps, in the slightlylater tower at Mocollop (Co. Waterford)which may have been built in direct hom-age (see below). Other significant differ-ences between Châteaudun and Pembrokesuggest that it was not a leading influenceon Marshal: the internal proportions arevery different, while Châteaudun showsneither a basal batter nor hourding-sock-ets, while containing both a well (housedin a masonry shaft) and a mural passage,neither of which occur at Pembroke. It ispossible, moreover, that it was an integrat-ed chamber-tower: toothing in the face-



Fig. 29 – The main entrance at Monkton Priory church, Pembroke, from c. 1200. The doorwayat Pembroke St Mary is very similar, as perhaps was the entry into Pembroke’s donjon.

Fig. 30: The roofing/dome support arrangement in the donjons at (L-R) Conisbrough and ChâteauGaillard. A similar ring of corbels can be seen at both (Conisbrough from Brindle and Sadraei 2015).



work suggests that a contemporarybuilding was formerly attached (Corvisier1998(2), 148-58; Marshall 2002b, 34). Instead, an immediate and perhaps morepersuasive antecedent for Pembroke’sdome may lie closer to home, in Conis-brough’s donjon. The upper works herehave long been interpreted as an attic floor(see Brindle and Sadraei 2015, 9, 13),which sits rather uneasily within the strati-fied social environment of the donjon,83

and Philip Davis’s suggestion that a timberdome instead crowned the tower is bothcompelling and consistent with the struc-tural evidence (Davis 2016, 254-7).84 Heenvisaged a high-status third-floor cham-ber beneath a grand, decorative domedceiling – precisely the kind of scheme thatwe can picture at Pembroke which may bethe realisation, in stone, of the same model(see Appendix 1). The Pembroke masterwas apparently familiar with, and probablyhad experience of the ecclesiastical archi-tectural tradition of Angevin Anjou andAquitaine, and its masonry domes at, forexample, the royal funerary church at Fon-tevraud Abbey, Maine-et-Loire (Bony1980, 86; Hulme 2014, 220). The structural evidence suggests that asimilar arrangement to that at Conisbroughexisted in the donjon at Château Gaillardwhich, at summit level, shows the samecircle of closely-spaced internal corbels (asnoted by Hislop 2016, 39; Fig. 30). A con-centric inner ‘parapet’ wall rose above thewall-walk at Conisbrough (Brindle andSadraei 2015, 9), arguably carrying a coni-cal roof over the timber dome (Davis 2016,251, 254); similar walling has been suggest-ed at Château Gaillard (Hislop 2016, 147;Fig. 31), and may be paralleled at Pembroke(see Appendix 1). The upper chamber atChâteau Gaillard, which is the entrance

floor, has been interpreted as an anteroomserving an audience chamber below (Mar-shall 2002b, 32); the large window over-looking the river suggests it may also havebeen a prospect chamber, consistent with agrand timber dome above. The Conisbrough donjon was built in the1180s, Château Gaillard in 1196-8, andPembroke in 1201-7 – around a twenty-year period: it is possible that successivemasons from the same Norman/Angevintradition worked at all three.85 Each showsa pronounced basal batter which, at Châ-teau Gaillard, rises two-thirds the height ofthe tower, possibly due to its location nearthe edge of a chalk cliff. Conisbrough andPembroke are moreover among the fewcylindrical donjons with four storeys. Noclose personal association between Marshaland Hamelin of Anjou is known,86 but theymoved within the same court ambit whileboth – respectively as favourite, and nota-bly loyal blood-relative – had close contactwith the Angevin kings whose generouspatronage, until 1207, Marshal enjoyed. The three donjons share the same spatialrelationship with the domestic buildings intheir respective castles: it is close, but notintegrated. Differences in their appoint-ments however suggest differing uses.Conisbrough contains fireplaces, latrines,basins and a private chapel; Pembrokeonly shows fireplaces (see Appendix 1)while Château Gaillard lacks both latrinesand any fireplace. Pembroke and Conis-brough feature hourding-sockets (Brindleand Sadraei 2015, 13; see Appendix 1),but not Château Gaillard. A large cisternlies beneath basement level at ChâteauGaillard (see Renn 1968, 38), and a well atConisbrough (Brindle and Sadraei 2015,8), but neither is so far known at Pem-broke. Curving mural stairs were em-



ployed at Conisbrough and a spiral stair atPembroke, while Château Gaillard appearsto have been served by internal timberstairs.87 A fine, white Magnesian limestone,ideal for closely-jointed ashlar faceworkand similar to that used at Château Gaillard,Chinon, Loches etc., was available locallyat Conisbrough, allowing its Angevin asso-ciations to be emphasised. No such materialwas on hand at Pembroke, which is rubble-built from the same tough Carboniferouslimestone as the cliff on which the castlestands; in this, however, it has precedents inthe undressed facework used at Château-sur-Epte and Neaufles (Fig. 2). There are however further similaritiesbetween the three castles, which may bemore than incidental. The comparable rela-tionships between donjon and curtain wallhave been discussed above. The curtain atConisbrough, which may be from the late1190s (Guy 2016a, 151), has solid semicir-cular interval turrets as at Henry II’s Lochesand Rouen (see above), and like those in hisdonjon at Niort (Mesqui 1997, 266-7; Mes-qui 2013, 196); these turrets may also bereflected in the lobate inner curtain, other-wise unparalleled, at Château Gaillard (His-lop 2016, 147; Tietzsch-Tyler 2013, 154)and in the solid semicircular turret on theWogan screen wall at Pembroke. In contrast to the above-mentioned towers,it is the rib-vault that predominates in KingPhilip’s Tours Philippiennes (Mesqui 2013,215-6, Fig. 7). And it is often used at everylevel, sometimes in combination with bar-rel-vaulting, to give summit-vaults at manysites e.g. Lillebonne and Gisors, but withflat surfaces as at Châteaudun. Its absencefrom Pembroke’s donjon is another attributethat sits uneasily with suggested Philippi-enne influence. And though the rib-vaultbecame widely adopted in cylindrical great

towers elsewhere in France it was, likebarrel-vaulting, seldom similarly used inthe British Isles.88

Nevertheless, the Pembroke dome mayrepresent the earliest secular use of anykind of vaulting in west Wales and, like somany of the castle’s attributes, influencedregional building, secular and ecclesiasti-cal, throughout the thirteenth century andbeyond (see above).89 The dome was emu-lated by summit-vaults in later work atPembroke Castle itself, in the engaged,dominant tour-mixte built in c.1230 at thesoutheast corner of nearby Manorbier Cas-tle (King and Perks 1970, 107, 114-15) –caput of a subordinate barony of Pem-broke – and in the similar northwest towerat Laugharne Castle, Carms., from the1240s-50s (Avent 1995, 9-10, 34). Thesouthwest tower at Kidwelly, from the1270s, has a flatter saucer vault at thesummit, to which a second was addedwhen the tower was heightened in theearly fourteenth century (Kenyon 2007,9-11, 37), while the slate roofs over bothtowers at Younger Marshal Cilgerran mayhave mimicked shallow domes (Hilling2000, 15 fig.).90 Similarly, the twin-lightdonjon windows at Pembroke were emu-lated in the engaged, dominant easterntour-mixte at Cilgerran, in the similarsoutheast tower at Manorbier, and in laterwork at Pembroke itself, in both the castleand the town walls.91

Such emulation shows ‘the admiration inwhich Pembroke was held’ (Goodall 2011,162-3; also see Wiles 2014), but also rep-resented a visible means by which Marshaltenant lords might express their homage to– and affinity with – the earls. The sameprocess may also have been at work inIreland. Mocollop Castle, Co. Waterford,shows a free-standing cylindrical donjon



brough from Brindle and Sadraei 2015).Fig. 31: ABOVE: Viollet-le-Duc’s reconstructionof the donjon at Château Gaillard.

Fig. 33 ABOVE: Plans of the ‘towered keeps’ at(top) Nemours, Seine-et-Marne, and (bottom)Carlow Castle (from Leask 1964, and Mesqui2013).


Fig. 32. BELOW: First-floor plan of the donjonat Le Châtelier (Angevin Touraine), showingthe stair turret. The curtain may be closelycontemporary (from Corvisier 1998).


begun 1215-20 (O’Keeffe and Whelanforthcoming; Tietzsch-Tyler 2018, 121). Itwas almost certainly built by Thomas Fitz-Anthony (d. 1226/7), who had been grantedthe castle and lordship in 1215 (Sweetman1875, 89-90). The window(s) may emulatethose at Pembroke, while the third storeyshows a vault which O’Keeffe and Whelanconsider to have been built when the towerwas heightened in the fifteenth century.However, FitzAnthony had been part of theMarshal mesnie since 1207, and was theirseneschal of Leinster 1215-25 (Bradley andMurtagh 2017, 242; Crouch 2015, 89-93;Holden et al. 2004, 177), and it has beensuggested that the feature is primary, builtby FitzAnthony as a summit-vault in emula-tion of his lord’s donjon at Pembroke (DanTietzsch-Tyler, pers. comm.).The cylindrical donjon in Wales andIrelandPembroke may be the earliest free-stand-ing, fully-circular donjon in the BritishIsles: the rest appear to be somewhat later,enjoying their heyday in the 1220s-30s.However, the similar tower at Dundrum,Co. Down, was begun by Hugh de Lacybetween 1205 and 1210, to celebrate hiselevation to the earldom of Ulster; minorworks by King John in 1210-11 representits completion (McDonald 2017, 270-1,276; McNeill 2003, 98, 105). Thoughsmaller overall, and comprising just threestoreys, like Pembroke it is a free-standingcylinder without projections, on levelground, with a basal batter and a spiral stairthat climbs the full height of the tower; italso shows evidence for a timber porch, asat Pembroke (see Appendix 1; Guy 2015,50). There is however neither dome norexternal offsets, architectural detail is ab-sent, windows appear to be single lights(while the basement is pierced by slit-

lights), and a large cistern lay beneathbasement level as at Château Gaillard. The Lacy brothers, Hugh of Ulster andWalter, lord of Meath, were close associ-ates of William Marshal with whom theywere military allies in 1207-8, perhapsbelonging to his wider affinity (Crouch2016, 124-30, 218; Holden et al. 2004,189). It is possible therefore that Pem-broke, which was probably more-or-lesscomplete when Marshal left for Ireland in1207, was an influence at Dundrum. But,while Hugh de Lacy is not recorded inFrance until his exile of 1210-1219/21(Duffy 2017, 296; Smith 2006), Walterwas part of King John’s entourage in Nor-mandy from September 1199 until March1201: he witnessed deeds at Rouen in Sep-tember 1199 (Hardy 1837, 24) and at Caenand Falaise in June 1200 (Hardy 1837, 67,69), while he presumably accompanied theking in Poitou and Touraine during thisperiod. It is possible, then, that Dundrumwas instead, or additionally inspired by theAngevin circular donjons of these regions. Walter de Lacy also provides a linkbetween Ireland and the cylindrical don-jons of southeast Wales and the Marches,of which one of the earliest may be histower at Longtown Castle, Herefs., almostcertainly built between his return fromexile in 1213 and his departure for Irelandin 1224 (Duffy 2017, 316; Hislop 2016,125):92 Walter was sheriff of Herefordbetween 1215 and 1223, spending most ofthis period in the region (Flanagan 2004).The Longtown donjon comprises threestoreys, like Dundrum, while structuralevidence indicates that both were flooredusing radial cross-joists (McNeill 2003,104). Unlike Dundrum, however, it occu-pies a motte. It also features three project-ing, semicircular turrets, one of them



housing the spiral stair in a manner thatbecame emblematic of cylindrical donjonsin the region. Though absent from Dun-drum and Pembroke, a semicircular stair-turret had been employed in AngevinFrance in the cylindrical donjon, probablyfrom 1180-1200, at Le Châtelier (Corvisier1998(2), 185-8; Fig. 32) the solid turretsrecall those at Conisbrough, Loches andNiort (see above), and, perhaps, the semi-circular pilasters in the round tower at Con-dé-sur-Noireau (Calvados) in Normandy,built under Angevin rule 1170-89 (Corvisi-er 1998(2), 256-62).93 While Lacy influ-ence may lie behind the wider adoption ofthe semicircular stair turret in Irish cylin-drical towers of the mid thirteenth-century,e.g. Kiltinane, Co. Tipperary (see McNeill1997, 94-5), it may already have been em-ployed by King John in the tour-mixte atArdfinnan (also Co. Tipperary);94 the simi-lar turret against the circular northeast tow-er at Helmsley Castle, built 1191-1227 byRobert de Roos, was probably also influ-enced by royal precedents (see above; Ke-nyon 2017, 23). Lyonshall Castle, (Herefs.), near Long-town, was held of Walter de Lacy’s honorof Weobley. It features the scant remains ofa free-standing cylindrical donjon (Guy2017, 119-21; Renn 1961, 141), which iswithout a turret and may therefore beamong the earliest in the region. Its tenantlord, and builder, was the Marshals’ house-hold knight and kinsman Stephend’Evreux, who is recorded in their affinityfrom 1199 until his death in 1228, and whorepresented the elder Marshal, in Leinster,during the Lacy alliance of 1207-8 (Holdenet al. 2004, 175; 2006, 153). It is difficultto imagine that it was conceived indepen-dently of the Pembroke donjon, but directinfluence from Angevin France appears tolie behind the presence of a surrounding

‘chemise’ wall like those around the tow-ers at Château-sur-Epte, Neaufles, andmany others including Condé-sur-Noireauand Fréteval where, as at Lyonshall, thereis no motte. In many cylindrical donjons nearby, forexample Caldicot and Skenfrith (bothMon.), the semi-circular stair turret occursalongside a second feature that is equallycharacteristic of the region, but absent fromAngevin France, Ireland, and the Pembrokedonjon: a roll-moulded basal string-course(Guy 2009, 23, 82-107; Renn 1961, 132).There is also a string-course at Lyonshallbut it is simply-chamfered (Renn 1961,141), perhaps confirming that the donjon isearly and, along with Longtown, a primaryinfluence in the region.95 Nevertheless, Cal-dicot and Skenfrith were built by patronswith extensive experience of Normandy,Touraine and Poitou, and possibly receivedat least some inspiration directly from An-gevin examples: Caldicot was possibly be-gun shortly after Longtown by Henry deBohun (d. 1220), who had enjoyed a cross-channel lifestyle until the loss of Normandyin 1204 (Powicke 1913, 487-8) – althoughit may be later – while Hubert de Burgh,builder at Skenfrith, was a senior Angevincommander in France and former sene-schal of Poitou.96

Influence from Pembroke’s donjon was, inwest Wales, mainly in matters of detail,discussed above; surprisingly, perhaps, noother cylindrical donjon is known in itsimmediate locality, although its visual pres-ence is evoked in the towers at YoungerMarshal Cilgerran, to which homage waspaid at Manorbier and Laugharne. The twocircular donjons of west Wales, at Dinefwrand Dryslwyn, Carms. – both c.1230s (Reesand Caple 2007, 10) – show basal string-courses and were clearly influenced fromsoutheast Wales. The base of a very small



round tower on the motte at Nevern Castle,Pembs., is probably too small to be a func-tional donjon (nine metres in diameter)while its masonry was bonded with clay(Caple 2016, 387); it may represent a purelysymbolic tower, possibly uninhabitable andnot closely related to the Pembroke donjon,of the kind discussed by Pamela Marshall(Marshall 2002b, 27, 34). While the excava-tor favours a date before 1195 for all mason-ry at the castle (Caple 2016, 384),compelling arguments have been made thatthe tower was an addition of 1204-15 (Wiles2014, 189; also see Turvey 1989, 64).97

The remaining cylindrical great towers ofthe British Isles appear to be from the1220s-40s, or later, and similarly resultfrom secondary, internal influences (see,inter alia, Goodall 2011, 181; Renn 1961,135). In Ireland many are engaged astours-mixtes or chamber-towers (McNeill1997, 95), probably influenced by Dublinand Kilkenny, e.g. Nenagh and Kiltinane,both Co. Tipperary (see McNeill 1997, 31,88, 94; Murtagh 2017, 162-75), while theLacy’s free-standing tower at Clogh Ough-ter, Co. Cavan, follows their lead at Dun-drum. But we have seen above thatMocollop may have been directly influ-enced from Pembroke.Philippienne influence: the youngerMarshalsThe ‘towered keeps’ of Ireland, concen-trated in Leinster (see Leask 1964, 143;McNeill 1997, 118-24; Sweetman 1999,39, 60; et al.), are of a form unique in theBritish Isles until the fourteenth century,when a similar plan was adopted at e.g.Dudley (Worcs.) and Nunney (Somerset).In the absence of any indigenous anteced-ents in Ireland, it is thought that theirorigins lie in similar towers built inFrance (see O’Keeffe 1990, 22, et al.).

We have seen that solid, circular turretsappear at Henry II’s Niort and, also inthe Angevin domain, the donjon at Châ-teau Romefort, Ciron (Indre, in Tou-raine), from c.1195-1200 (Mesqui 2013,207). Hollow circular towers were usedat Vic-sur-Aisne (Aisne, in Picardy),which was probably built during the1190s (Rolland 1984, 143), but therewere only two, at one end of the donjon. The fully-developed plan, with fourhollow corner towers as in Ireland,seems first to appear at Nemours (Seine-et-Marne), built by the Capetians’ highchamberlain around 1200 (Mesqui 2013,128 and fig., 207; Fig. 33). Given itsdate, and its location in the Île-de-France south of Paris, it cannot havebeen seen by William Marshal I (or hismasons) before – nor probably during –his French visit of spring 1216, afterwhich work on his Irish castles is in anycase unlikely (see above).98

Richard Marshal however appears tohave visited Nemours: he was retained,from late 1217 until mid-1219, in thehousehold of King Philip who is knownto have stayed at the castle during thisperiod (Crouch 2015, 23, 190-3; Power2003, 210, 213-6; Samaran and Nortier1979, 127-9).99 Richard succeeded to hisfather’s Norman lands in late summer1219, spending most of his time thereuntil 1231. Yet he maintained regularcontact with his brother William Mar-shal II during the thaw in Anglo-Frenchrelations of the early 1220s, when peri-odic visits by Richard, to both Britainand Ireland, are suggested in the sources(Holden et al. 2006, 32; Power 2003,225 n.).100



The towered keeps at Ferns (Co. Wexford)and Lea (Co. Laois) are now thought to befrom the mid-thirteenth century (Dempsey2017, 247; McNeill 1997, 144; O’Keeffeand Coughlan 2003, 144-8).101 That at Car-low is suggested to have been begun by theelder Marshal, 1210-15 (O’Conor 1997, 15-16; Tietzsch-Tyler 2018, 121; Fig. 33), butthe evidence presented above, taken with itscruciform loops, argues that the toweredkeep concept is more likely to have beenintroduced to Leinster, ultimately fromFrance, by his sons during the 1220s.102 Itmay be significant that William Marshal IIwas quit of a years’ service to the king inboth 1222 and 1225 ‘in aid of the fortifica-tion of one of his castles in Ireland’ (Hardy1844a, 32; Sweetman 1875, 158, 192),103

while the town of Carlow received its firstcharter from the earl in 1223 (Ballard andTait 1923, xxxv, 16, 53). The younger Marshals may provide acontext for other Philippienne attributesseen in Ireland in the 1220s, such as theclockwise/counter-clockwise stair systemin the donjon at Mocollop, which has prec-edents at e.g. the Tour du Coudray at Chi-non (O’Keeffe and Whelan forthcoming);the same French influence is implicit in thecylindrical lighthouse tower at Hook Point,Co. Wexford, begun by Walter Marshal in1241-5, which is rib-vaulted at every levellike a Tour Philippienne (Murtagh 2016,230; see Appendix 2). While he was withPhilip Augustus 1217-19, Richard Marshalwill have seen a number of Capetian cas-tles: the king was frequently at both theLouvre and Gisors, as well as Péronne(Somme), Vernon (Eure) and Montargis(Samaran and Nortier 1979, 126-7, 185-6,240-6), all castles with Philippienne plan-ning and/or Tours Philippiennes, and allthought to have been underway by 1217(Mesqui 1997, 457, 464; Mesqui 2013,

42-3, 162); Richard is specifically record-ed at Montargis after his father’s death inJuly 1219 (Holden et al. 2006, 194). Inaddition, William Marshal II and Richardtogether visited King Philip in June 1220,spending time at Paris and Melun (Crouch2015, 252-4, 290-91; Stapleton 1844,cxxxviii), after which Richard maintainedhis close associations with the Frenchcourt (Crouch 2015, 23; Power 2003, 213-16). William’s continued contact with hisbrother, and with France, is manifest in hisrecruitment in 1224 of an author, fromTouraine or Anjou, to compose the bio-graphical Histoire (Holden et al. 2006,22).104

Philippienne influence may also be ap-parent in Pembroke’s Dungeon Tower,which is clearly secondary to the elderMarshal’s inner curtain of c.1204-7 (King1978, 105, 118).105 Neil Guy has observedthat the exterior shows three deep sockets,following a spiral incline like the helicalscaffold sockets seen to great effect inEdward I’s North Wales castles (Guy2016b, 30 and appendix; Fig. 34). Helicalscaffolding appears to have been first usedin King Philip’s castles, c.1200-1223 (in-cluding Dourdan, Rouen, Villeneuve-sur-Yonne and Verneuil), but afterwards waschiefly confined to Savoy and NorthWales (Hislop 2016, 60, 119). While it ispossible that the Dungeon Tower (Fig. 34)was added by the elder Marshal, followinghis visit to King Philip in spring 1216(when the sources may hint at attention towest Wales: Hardy 1833, 228, 271), wehave seen that his movements do not ap-pear to have taken in any Philippiennecastles. The Dungeon Tower shows manyMarshal attributes including rounded arch-es, entasis, a summit vault and narrow-splayed loops – some of which are cruci-form – but all appear in Younger Marshal




Fig. 34 – The Dungeon Tower at Pembroke Castle. The threeinclined sockets can be clearly seen.

Fig. 35 – Sketch elevation of the donjon and Dungeon Tower at PembrokeCastle, from southeast, showing relative levels.


work elsewhere while, in contrast to thedonjon, dressed stone is more-or-less ab-sent. Externally, it is confined to two basaloillets, one rectangular, the other circular,and while both are probably medieval – thelatter is shown in pre-restoration photo-graphs – they may be secondary insertions(see above); the suspicion may be borne outby their anomalous oolite dressings. Asnoted above, the tower’s affinities appear tolie more closely with younger Marshalwork at e.g. Cilgerran and Chepstow. Thesimple, annular ‘capital’ that terminates thespiral-stair newel is also in a style morecharacteristic of the 1220s-30s and may bethe work of a mason with experience of theWest Country School, in which similar ter-minal capitals appear in imposts and verti-cal roll-mouldings (Brakspear 1931, 8-9);the school retained its influence in south-west England, south Wales and the westMidlands into the 1240s, and lies behind thedetail in younger Marshal work at Chep-stow, chiefly by Gilbert Marshal 1234-41(Coldstream and Morris 2006, 103, 112). Of the younger Marshals, William II maybe the most likely builder of the DungeonTower, under French influence via hisbrother Richard; the latter’s activities asearl, 1231-4, suggest he was not personallyresponsible. William appears to have beenin Pembrokeshire in 1221 (Jones 1952, 98),and had been commanded to ‘repair’ thecastles and Narberth and Wiston (Pembs.)in October 1220, following Welsh attacks(Pat. Rolls 1216-25, 254-5). He was inwestern Wales, on campaign, from April-October 1223 (Crouch 2015, 18-19, 285-9;Edwards 1935, 3-4, 30), and again in Au-gust 1225, when there is some suggestion inthe sources that he was strengthening hiscastles (Crouch 2015, 195-6; Hardy 1833,59, 79). William headed a delegation tomeet Llywelyn ap Iorwerth later in that year

(Crouch 2015, 20), perhaps significantlyas the earliest use of helical scaffoldingotherwise known in Britain is in the cylin-drical donjon at Llywelyn’s DolbadarnCastle, Caerns., normally assigned to the1230s (Goodall 2011, 227); it has hithertobeen regarded as an isolated instance untilthe Edwardian castle-building of the1270s-80s.106 But Marshal influence atDolbadarn is in no sense a given and,while it is difficult to envisage anotherconduit for Llywelyn’s use of such scaf-folding, his personal contact with the earlswas limited (Walker 1994, 43-4); there aremoreover few other notably ‘Marshal’ at-tributes in the Dolbadarn donjon, apartfrom its circular shape.107

ConclusionAlmost all features of William Marshal I’scastles have precedents in works of theEnglish Crown. Marshal was unusualamong major barons in that he owed ev-erything to his royal patrons (later, thesame could be said of Hubert de Burgh),and he had no inheritance of his own untilhis brother died in 1194. So it shouldcome as no surprise that Marshal chose toemulate Angevin buildings: it can be ar-gued that his references to them were de-liberate, and intended to be understood.Marshal’s emulation may also have beenanother expression of the loyalty that hewas so keen on proclaiming, and was per-haps facilitated by the loan of royal ma-sons. His debt to the French Crown was oflesser magnitude and, politically, some-thing to be played down. While other in-fluences may be present in his buildings,there is no evidence that he drew anydirect inspiration from Philip Augustus’sworks, nor of a mechanism through whichit could be transmitted. Philippienne workhas little in common with the earl’s and is



mainly later than 1200, a period when hos-tilities prevented constructive Anglo-French contact. Neither was Philip a greatbuilder of donjons, generally preferring thedefensive tour-mixte. The evidence discussed above suggests thatMarshal may not have begun any seriousbuilding in the British Isles until after he wascreated earl in 1199; it is only then that hereveals an interest that, hitherto, had largelybeen confined to his estates in Normandywhere he was resident during the later 1190s.It is possible that the Pembroke donjon, be-gun in 1201, was his first major British con-struction, marking his elevation to theearldom at its titular caput; his main castle atChepstow already had an impressive greattower, built in the eleventh century. Both architecturally and functionally, thePembroke donjon belongs to an Angevintradition well-established in northernFrance. Marshal presumably engaged hisPembroke team in late 1200 or early 1201.Full-scale war the following year may haveplaced a demand on experienced masons, forNormandy, but there is no evidence for ahiatus in the lower stages of the Pembrokedonjon. Work similarly continued uninter-rupted at King John’s Corfe Castle where amajor campaign took place from 1201 to1204, including the Gloriette in which webegin to see an identifiably ‘English’ form ofearly Gothic (Goodall 2011, 159, et al.),emulated by Marshal at Chepstow (seeabove). The Pembroke master also refer-enced Angevin precedents and may evenhave been instructed, by Marshal, to followa particular template, cf. the contract from1224 in which Count Robert III of Dreuxinstructs his master mason to build a donjonspecifically like the one at Nogent-le-Rotrou, Eure-et-Loir (reproduced in Jost2002, 182-3). Such contracts, which became

more frequent during the later Middle Ag-es, show that patrons could, from an earlydate, be responsible (at least in part) foroverall design, as suggested of Richard I atChâteau Gaillard (Hislop 2016, 148; Mes-qui 1993, 334, et al.; also see Hulme 2014,230). The Pembroke team seems to haveincluded free-masons from the West Coun-try, stylistic evidence for which is mostapparent in the lower two storeys; after1204, at least, masons who had personallyworked in Angevin France appear also tohave been employed and their influence isfelt in the upper storeys of the donjon,including the windows and dome. Theywere also responsible for the similar win-dow in the Wogan cavern. Marshal’s work elsewhere, which maylargely belong to the post-1204 period, sim-ilarly followed Angevin practice. It was notuntil his sons succeeded in 1219, with theirclose associations with the Philippiennecourt, that French influence began to be feltin Marshal design. However, it remainedsecondary to their established practices.

A wider debate concerns the extent ofCapetian influence on the Angevin kings,as well as their barons. The Angevinarchitectural tradition was very broad-based (see Gillingham 2001, 118) and, incommon with much of western Europe,reveals the dominating influence of theFrench Gothic aesthetic. But Capetianinfluence must not be taken as a given.The two realms were subject to the sametrends in military design, and discerningspecific patterns of influence is not alwayspossible – while, as French authors arequick to point out, the current could de-monstrably flow the other way. It has beenseen above that many aspects of Angevindesign originated independently ofFrench precedent.



AcknowledgementsMany thanks to David Crouch, the late PhilipDavis, Neil Guy, Malcolm Hislop, Jacques leMaho, Jean Mesqui, Denys Pringle, DanTietzsch-Tyler and the late Rick Turner forinformation, images and discussion.

ReferencesPrimary sources (published)Ballard, A. and Tait, J. (eds), 1923 BritishBorough Charters 1216-1307 (CambridgeUniversity Press). Brewer, J. S. (ed.), 1863 Giraldi Cambren-sis Opera, 3: ‘De Invectionibus’, ‘De Me-nevensi Ecclesia Dialogus’ and ‘Vita S.David’ (London: Rolls Series).Calendar of Charter Rolls 1, Hen. III 1226-1257 (London: HMSO, 1908).Calendar of Charter Rolls 2, Hen. III 1257-1300 (London: HMSO, 1906)Caley, J., Ellis, H. and Bandinel, B. (eds),1846 Monasticon Anglicanum, by Sir Wil-liam Dugdale, 4 (London: Record Commis-sion).Crouch, D. (ed.), 2015 The Acts and Lettersof the Marshal Family: Marshals of Englandand Earls of Pembroke, 1145-1248 (Cam-bridge University Press: Camden 5/47).Davies, J. C. (ed.), 1946 Episcopal Actsrelating to the Welsh Dioceses 1066-1272,1 (Cardiff: Hist. Soc. Church in Wales 1).Edwards, J. G. (ed.), 1935 Calendar ofAncient Correspondence concerning Wales(Cardiff: University of Wales Press).Hardy, T. D. (ed.), 1833 Rotuli LitterarumClausarum 1, 1204-1224 (London: RecordCommission).Hardy, T. D. (ed.), 1835a Rotuli LitterarumPatentium 1/1, 1201-1216 (London: Re-cord Commission).

Hardy, T. D. (ed.), 1835b Rotuli Norman-niae, Johanne et Henrico Quinto, AngliaeRegibus, 1 (London: Record Commission).Hardy, T. D. (ed.), 1837 Rotuli Chartarum1/1, 1199-1216 (London: Record Com-mission).Hardy, T. D. (ed.), 1844a Rotuli Litterar-um Clausarum 2, 1224-1227 (London:Record Commission).Hardy, T. D. (ed.), 1844b Rotuli de Liber-ate ac de Misis et Praestitis, regnanteJohanne (London: Record Commission).Holden, A. J., Gregory S. and Crouch, D.(eds), 2002 History of William Marshal, 1(London: Anglo-Norman Text Society 4).Holden, A. J., Gregory S. and Crouch, D.(eds), 2004 History of William Marshal, 2(London: Anglo-Norman Text Society 5).Holden, A. J., Gregory S. and Crouch, D.(eds), 2006 History of William Marshal, 3(London: Anglo-Norman Text Society 6).Howlett, R. (ed.), 1889 ‘The Chronicle ofRobert of Torigni’, in Chronicles, Ste-phen, Henry II and Richard I, 4 (London:Rolls Series).Huygens, R. B. C. (ed.), 1976 MonumentaVizeliacensia: textes relatifs à l’histoirede l’abbaye de Vézelay (Turnhout:Brepols).Jones, T. (ed.), 1952 Brut y Tywysogyon:Peniarth MS. 20 Version (Cardiff: Univer-sity of Wales Press).Landon, L., 1935 The Itinerary of KingRichard I, with studies on certain mattersof interest connected with his reign (Lon-don: Pipe Roll Society 51).Packard, S. R. (ed.), 1927 MiscellaneousRecords of the Norman Exchequer, 1199-1204 (Northampton, Mass.: Smith Col-lege Studies in History 12, 1-4).

William Marshal, Pembroke Castle and Angevin design - References


Patent Rolls, Hen. III 1216-1225 (London:HMSO, 1901).Patent Rolls, Hen. III 1225-1232 (London:HMSO, 1903).Round, J. H. (ed.), 1899 Calendar of Docu-ments Preserved in France 1, AD 918-1206(London: HMSO).Samaran, C. and Nortier, M. (eds.), 1979Recueil des actes de Philippe Auguste Roide France, 4, 1 Novembre 1215 – 14 Juillet1223 (Paris: Imprimerie Nationale).Stapleton, T. (ed.), 1840 Magni RotuliScaccarii Normanniae 1 (London: Soc. ofAntiquaries).Stapleton, T. (ed.), 1844 Magni RotuliScaccarii Normanniae 2 (London: Soc. ofAntiquaries).Sweetman, H. S. (ed.), 1875 Calendar ofDocuments relating to Ireland, 1171-1251(London; HMSO).

Secondary sources (unpublished)Corvisier, C., 1998 ‘Les Grosses Tours deplan circulaire ou centré en France avant1200: étude sur les antécédents de la poli-tique castrale de Philippe Auguste’, 3 vols(PhD thesis: University of Paris).Day, A. and Ludlow, N., 2016 ‘PembrokeCastle Geophysical Survey 2016’ (DATArchaeological Services for Castle StudiesTrust – see:http:/castlestudiestrust.org/docs/Pembroke_Castle_ Geophysical%20_Survey_FINAL.pdf).

Guy, N. (ed.), 2016b ‘Summary notes forCSG study day at Pembroke’ (unpublishedhandout).Launay, A., 2015 ‘‘Qui a bâti les fortifica-tions de Pont-de-l’Arche?’(http://pontdelarche.overblog.com/article-qui-a-bati-les-fortifications-de-pont-de-l-arche-108299939.html).

Secondary sources (published)Ashbee, J., 2009 Goodrich Castle (Lon-don: English Heritage).Avent, R., 1995 Laugharne Castle (Car-diff: Cadw).Avent, R., 2003 ‘William Marshal’sbuilding works at Chepstow Castle, Mon-mouthshire, 11891219’, in J. R. Kenyonand K. O’Conor (eds), 50-71.Avent, R., 2006 ‘William Marshal’s castleat Chepstow and its place in military ar-chitecture’, in R. Turner and A. Johnson(eds), 81-90.Avent, R. and Miles, D., 2006 ‘The MainGatehouse’, in R. Turner and A. Johnson(eds), 51-62.Avent, R. and Turner, R., 2006 ‘The Mid-dle Bailey’, in R. Turner and A. Johnson(eds), 63-70.Baudry, M-P., 2002 ‘La politique de forti-fication des Plantagenêts en Poitou, 1154-1242’, in J. Gillingham (ed.), 43-69.Biller, T., 1998 Die Adelsburg inDeutschland: Entstehung - Gestalt - Be-deutung (Munich: Deutscher Kunstverlag).Bony, J. (ed.), 1980 The Art of the West inthe Middle Ages, by Henri Focillon, 1:Romanesque Art (Oxford: Phaidon).Boüard, M. de, 1966 ‘Fouilles au châteaude Bonneville-sur-Touques’, Annales deNormandie 16/4, 351-78.Bradley, J. and Murtagh, B., 2017 ‘Wil-liam Marshal’s charter to Kilkenny, 1207:background, dating and witnesses’, in J.Bradley, C. Ó Drisceoil and M. Potterton(eds), 201-48.Bradley, J., Ó Drisceoil, C. and Potterton,M. (eds), 2017 William Marshal and Ire-land (Dublin: Four Courts Press).



Brakspear, H., 1931 ‘A West CountrySchool of Masons’, Archaeologia 81, 1-18.Brindle, S., 2015 Dover Castle (London:English Heritage).Brindle, S. and Sadraei, A., 2015 Conis-brough Castle (London: English Heritage).Brown, R. A., 1976 English Castles (Lon-don: Batsford).Caple, C., 2016 ‘Nevern Castle 2008 –2015: closing in on the first Welsh stonecastle?’, Medieval Archaeol. 60/2, 382-91.Chapman, A., 2007 ‘The gatehouse of Pe-vensey Castle’, Sussex Archaeol. Collec-tions 145 (2007), 97–118.Châtelain, A., 1983 Châteaux Forts et Féo-dalité en Île de France du XIeme au XIIIeme

siècle (Nonette: Éditions Créer).Châtelain, A., 1991 ‘Recherche sur les châ-teaux de Philippe Auguste’, ArchéologieMédiévale 21, 115-61.Chollet, S. and Gousset, J-M., 2012 'Laval:nouvelle datation dendrochronologique dela tour maîtresse du château et de sonhourd', Bulletin Monumental 170/3, 261-2.Church, S. D., 2015 King John: England,Magna Carta and the Making of a Tyrant(London: McMillan).Coldstream, N. and Morris, R. K., 2006‘The Architecture and Decoration of theMarshals’ Great Tower’, in R. Turner andA. Johnson (eds), 101-12.Colfer, B., 2013 Wexford Castles: Land-scape, Context and Settlement (Cork Uni-versity Press).Colvin, H. M. (ed.), 1963 A History of theKing’s Works 2: The Middle Ages (London:HMSO).Corvisier, C., 2002 ‘La Grosse Tour du Roià Compiègne: premier donjon circulaire

capétien et archétype des tours-maitressesde Philippe Auguste’, Bulletin de la So-ciété Historique de Compiègne 38, 9-45.Crouch, D., 2016 William Marshal: ThirdEdition (Abingdon: Routledge).Curnow, P. E., 1980 ‘Some Develop-ments in Military Architecture c.1200: LeCoudray-Salbart’, in Brown, R. A. (ed.),Proceedings of the Battle Conference onAnglo-Norman Studies 2, 1979 (Wood-bridge: Boydell), 42-62.Davies, J. A., Riley, A., Levesque, J-M.and Lapiche, C. (eds), Castles and theAnglo-Norman World (Oxford and Phila-delphia: Oxbow Books).Davis, P., 2016 ‘Interpretations of theroof and upper works in the great tower ofConisbrough Castle’, Castle StudiesGroup Journ. 29, 250-62.Dempsey, K., 2017 ‘Lea Castle, Co. Lao-is: the story so far’, Castles Studies GroupJourn. 30, 237-52.Deschamps, P., 1972 ‘Forteresses de laFrance médiévale’, Journal des Savants,Année 1971/2, 106-18.Deyres, M. M., 1984 ‘Le donjon deChâtillon-sur-Indre’, Bulletin trimestrielde la Societé Archéologique de Touraine36, 364-77.Dixon, P., 2002 ‘The myth of the keep’,in G. Meirion-Jones, E. Impey and M.Jones (eds), 913.Dufaÿ, B., 2011 ‘Un nouveau regard surla forteresse de Chinon: 1er bilan de 7années de fouilles archéologiques’, in LesAmis du Vieux Chinon (eds), Chinon, LeDestin d’une Forteresse (Chinon: LesAmis du Vieux Chinon), 85-103.Duffy, P., 2017 ‘From Carcassonne toCarrickfergus: The Legacy of de Lacy’s



Crusade experience in Britain and Ireland’,in P. Duffy, T. O’Keeffe and J-M. Picard(eds), 295-328.Duffy, P., O’Keeffe, T. and Picard, J-M.(eds), 2017 From Carrickfergus to Carcas-sonne: The Epic Deeds of Hugh de Lacyduring the Albigensian Crusade (Turn-hout: Brepols).Emery, A., 2006 Greater Medieval Housesof England and Wales, 1300–1500: Vol. 3,Southern England (Cambridge UniversityPress).Erlande-Brandenburg, A., 1970 ‘Le donjonde Châteaudun’, Bulletin Monumental128/2, 141-2.Fichet de Clairfontaine, F., Mastrolorenzo,J. and Brown, R., 2016 ‘Le château deFalaise (Calvados): état des connaissancessur l’evolution du site castral du dixièmesiècle au treizième siècle’, in J. A. Davies,A. Riley, J-M. Levesque and C. Lapiche(eds), 231-55.Flanagan, M. T., 2004 ‘Lacy, Walter de (d.1241), magnate’, Oxford Dictionary of Na-tional Biography (Oxford University Press:accessed 30 March 2018).Fleury, M. and Kruta, V., 2000 Le Châteaudu Louvre: guide et monographie (Dijon:Éditions Faton).Gillingham, J., 2001 The Angevin Empire(London: Hodder Arnold).Gillingham, J. (ed.), Anglo-Norman Stud-ies 24: Proceedings of the Battle Confer-ence, 2001 (Woodbridge: Boydell).Goodall, J., 2001 Richmond Castle/St Ag-atha’s Abbey, Easby (London: EnglishHeritage).Goodall, J., 2011 The English Castle 1066-1650 (New Haven and London: Yale Uni-versity Press).

Goodall, J., 2015 Warkworth Castle (Lon-don: English Heritage).Grant, L., 1994, ‘Le patronage architectur-al d’Henri II et de son entourage’, Cahiersde Civilisation Médiévale 37 (145-6), 73-84.Guy, N. (ed.), 2009 ‘Castle Studies GroupConference, Abergavenny’, Castle StudiesGroup Journ. 22, 4-127.Guy, N., 2013a ‘Conisbrough and Morte-mer Castles’, Castle Studies Group Journ.26, 203-6.Guy, N. (ed.), 2013b ‘Castle StudiesGroup Conference, ‘Castles of Co. Dur-ham’’, Castle Studies Group Journ. 26,4-150.Guy, N. (ed.), 2015 ‘Castle Studies GroupConference, ‘Castles of Northern Ireland”,Castle Studies Group Journ. 28, 4-119.Guy, N., 2016a ‘The portcullis – designand development – Chepstow Outer Gate’,Castle Studies Group Journ. 29, 192-201.Guy, N. (ed.), 2017 ‘Castle Studies GroupConference, ‘Castles of the Welsh March-es’’, Castle Studies Group Journ. 30, 4-150.Guy, N. (ed.), 2018 Castles: History, Ar-chaeology, Landscape, Architecture andSymbolism (Daventry: Castle StudiesGroup).Héliot, P., and Deyres, M., 1987 ‘Le châ-teau de Loches’, Bulletin Monumental145/1, 15-85.Heslop, T. A., 2003 'Orford Castle: nostal-gia and sophisticated living', in R. Liddi-ard (ed.), Anglo-Norman Castles(Woodbridge: Boydell), 273-296.Hilling, J., 2000 Cilgerran Castle/St Dog-maels Abbey (Cardiff: Cadw).



Hislop, M., 2016 Castle Builders: Ap-proaches to Castle Design and Construc-tion in the Middle Ages (Barnsley: Pen &Sword).Hislop, M., 2018 ‘The Round Tower ofBarnard Castle’, in N. Guy (ed.), 81-104.Hodkinson, B., 2003 ‘A summary of recentwork at the Rock of Dunamase, Co. Laois’,in J. R. Kenyon and K. O’Conor (eds),32-49.Hourihane, C. P. (ed.), 2012, The GroveEncyclopedia of Medieval Art and Archi-tecture, 2 (Oxford University Press).Hulme, R., 2014 ‘The Impact of ChâteauGaillard and Crusading on English CastleArchitecture’, Castle Studies Group Journ.27, 203-33.Impey, E. and Parnell, G., 2011 The Towerof London: The Official Illustrated History(London and New York: Merrell).Jost, B., 2002 ‘Meunzenberg (Hesse) andits relationship to later twelfth-century cas-tles’, in G. Meirion-Jones, E. Impey and M.Jones (eds), 179-87.Kennedy, H., 1994 Crusader Castles(Cambridge University Press).Kenyon, J. R., 1990 Medieval Fortifica-tions (Leicester University Press).Kenyon, J. R., 2007 Kidwelly Castle (Car-diff: Cadw).Kenyon, J. R., 2017 Helmsley Castle (Lon-don: English Heritage).Kenyon J. R. and O’Conor, K. (eds), 2003The Medieval Castle in Ireland and Wales:Essays in honour of Jeremy Knight (Dub-lin: Four Courts Press).King, D. J. C., 1977 ‘Pembroke Castle:Derivations and relationships of the domedvault of the donjon, and of the HorseshoeGate’ Château Gaillard 8, 159-69.

King, D. J. C., 1978 ‘Pembroke Castle’,Archaeol. Cambrensis 127, 75–121.King, D. J. C., 1983 Castellarium Angli-canum (New York: Kraus International).King, D. J. C., and Perks, J. C., 1970‘Manorbier Castle, Pembrokeshire’, Ar-chaeol. Cambrensis 119, 83-118.Knight, J. K., 1977 ‘Usk Castle and itsaffinities’, in M. R. Apted, R. Gilyard-Beer and A. D. Saunders (eds), AncientMonuments and their Interpretation: Es-says presented to A. J. Taylor (Chichester:Phillimore), 139-54.Knight, J. K., 1993 ‘Excavations at Mont-gomery Castle – Part 1’, Archaeol. Camb-rensis 141, 97180Knight, J. K., 2008 ‘Welsh Space andNorman Invaders: Usk Castle 1136-1245’,in J. K. Knight and A. Johnson (eds), UskCastle, Priory and Town (Almeley: Lo-gaston), 55-68.Knight, J. K., 2009 The Three Castles/HenGwrt Medieval Moated Site (Cardiff:Cadw).Knight, J. K., 2018 ‘Fog in the Channel?Influences in Anglo-French castle build-ing’, in N. Guy (ed.), 155-73.Langeuin, P., 2002 ‘Les campagnes deconstruction du château d'Arques-la-Ba-taille (XIe-XVe siècles)’, Bulletin Monu-mental 160/4, 345-378.Leask, H. G., 1964 Irish Castles and Cas-tellated Houses (Dundalk: DundalganPress).Lloyd, T., Orbach, J. and Scourfield, R.,2004 The Buildings of Wales: Pembroke-shire (New Haven and London: Yale Uni-versity Press).Ludlow, N., 1991 ‘Pembroke Castle andTown Walls’, Fortress 8, 25-30.



Ludlow, N., 2003 ‘The Castle and Lord-ship of Narberth’, Journ. Pembs. Hist. Soc.12, 5-43.Ludlow, N., 2014 Carmarthen Castle: theArchaeology of Government (Cardiff: Uni-versity of Wales Press).Ludlow, N., 2018 ‘Bothwell: a WelshMarches castle in Scotland?’, in N. Guy(ed.), 236-80.Manning, C., 2003 ‘The Record Tower,Dublin Castle’, in J. R. Kenyon and K.O’Conor (eds), 7295.Manning, C., 2013 Clogh Oughter Castle,Co. Cavan: Archaeology, History and Ar-chitecture (Dublin: Stationery Office).Marshall, P., 2002a ‘The CeremonialFunction of the Donjon in the Twelfth Cen-tury’, Château Gaillard 20, 141-51.Marshall, P., 2002b ‘The Great Tower asresidence in the territories of the Normanand Angevin kings of England’, in G.Meirion-Jones, E. Impey and M. Jones(eds), 27-44.McDonald, P., 2017 ‘Identifying Hugh IIde Lacy’s Contribution to Dundrum Castle(Co. Down)’, in P. Duffy, T. O’Keeffe andJ-M. Picard (eds), 263-76.McNeill, T. E., 1990 ‘Trim Castle, Co.Meath: the first three generations’, Archae-ol. Journ. 147, 308-36.McNeill, T., 1997 Castles in Ireland: Feu-dal Power in a Gaelic World (London andNew York: Routledge).McNeill, T. E., 2003 ‘Squaring circles:flooring round towers in Wales and Ire-land’, in J. R. Kenyon and K. O’Conor(eds), 96-106.McNeill, T. E., 2006 ‘The view from thetop’, in J. de Meulemeester (ed.), Mélangesd’archéologie médiévale: liber amicorum

en hommage à André Matthys (Namur:Les Cahiers de l’Urbanisme), 122-7.Meirion-Jones, G., Impey E. and Jones,M. (eds), 2002 The Seignurial Residencein Western Europe AD c800-1600 (Ox-ford: Archaeopress; British Archaeologi-cal Reports International Series S1088).Mesqui, J., 1993 Châteaux et Enceintes dela France Médiévale 2: la Résidence et lesÉléments d’Architecture (Paris: ÉditionsPicard).Mesqui, J., 1997 Châteaux forts et fortifi-cations en France (Paris: Flammarion).Mesqui, J., 2011 ‘La Tour des Archives etle Fort des Tourelles de Vernon (Eure)’,Bulletin Monumental 169/4, 291-318.Mesqui, J., 2013 Châteaux et Enceintes dela France Médiévale 1: les Organes de laDéfense (Paris: Éditions Picard).Mesqui, J., 2018 ‘Philippe Auguste‘tours”, in N. Guy (ed.), 129-139.Morris, R. K., 2010 Kenilworth Castle(London: English Heritage).Murtagh, B., 2016 ‘The medieval tower ofHook, Co. Wexford: A beacon of com-merce and symbol of Marshal power andwealth in Ireland’, Château Gaillard 27,223-236.Murtagh, B., 2017 ‘William Marshal’sGreat Tower at Pembroke, Wales: a viewfrom Ireland’, in J. Bradley, C. Ó Dris-ceoil and M. Potterton (eds), 111-82.O’Conor, K., 1997, ‘The Origins of Car-low Castle’, Archaeology Ireland 11/3,13-16.O’Keeffe, T., 1990, ‘The Archaeology ofNorman Castles in Ireland 2: Stone Cas-tles’, Archaeology Ireland 4/4, 20-22.



O’Keeffe, T., 2009 ‘Dublin Castle’s donjonin context’, in J. Bradley, A. J. Fletcher andA. Simms (eds), Dublin in the MedievalWorld: Studies in Honour of Howard B.Clarke (Dublin: Four Courts Press), 277-94.O’Keeffe, T., 2017 ‘Trim before 1224:New Thoughts on the caput of de LacyLordship in Ireland’, in P. Duffy, T.O’Keeffe and J-M. Picard (eds), 31-56.O’Keeffe, T. and Coughlan, M., 2003 ‘Thechronology and formal affinities of theFerns donjon, Co. Wexford’, in J. R. Ke-nyon and K. O’Conor (eds), 133-48.O’Keeffe, T. and Whelan, D., forthcoming‘A Little-Known Anglo-Norman Castle ofthe First Rank: Mocollop, Co. Waterford’,in B. Murtagh (ed.), Castles and defencesin Ireland and abroad: essays in honour ofDavid Newman Johnson.Page, W. (ed.), 1907 A History of theCounty of Leicestershire, 1 (London: Vic-toria County History).Painter, S., 1933 William Marshal: Knight-Errant, Baron, and Regent of England(Baltimore: Johns Hopkins Press).Pitte, D., 2002 ‘Château-Gaillard dans ladéfense de la Normandie orientale (1196-1204)’, in J. Gillingham (ed.), 163-75.Pitte, D., 2007 ‘Arques-la-Bataille, sond-age archéologique dans le château’, Bulle-tin Monumental 165/1, 113-14Power, D., 2003 ‘The French interests ofthe Marshal earls of Striguil and Pembroke,11891234’, in J. Gillingham (ed.), Anglo-Norman Studies 25: Proceedings of theBattle Conference, 2002 (Woodbridge:Boydell), 199-225.Powicke, F. M., 1913 The Loss of Normandy(1189-1204): Studies in the History of the An-gevin Empire (Manchester University Press).

Pringle, D. (ed.), 1988 Crusader Castlesby T. E. Lawrence (Oxford UniversityPress).Pringle, D., 1989 ‘Crusader Castles: TheFirst Generation’, Fortress 1, 14-25.Pringle, D., 2014a ‘Crusader Castles andFortifications: The Armenian Connec-tion’, in C. Mutafian (ed.), La Méditer-ranée des Arméniens, xiie–xve siècle (Paris:Geuthner), 353–72.Pringle, D., 2014b ‘Edward I, Castle-building and the Tower of the English inAcre’, in C. Donovan (ed.), A Fresh Ap-proach: Essays Presented to Colin Platt inCelebration of his Eightieth Birthday(Bristol: Trouser Press), 48-56.Radford, C. A. R., 1958 Goodrich Castle(London: HMSO).Rees, S. and Caple, C., 2007 DinefwrCastle/Dryslwyn Castle (Cardiff: Cadw).Remfry, P., 2011 ‘White Castle and theDating of the Towers’, Castle StudiesGroup Journ. 24, 213-26.Renn, D. F., 1961 ‘The Round Keeps ofthe Brecon Region’, Archaeol. Cambren-sis 110, 129-43.Renn, D. F., 1968 ‘The Donjon at Pem-broke Castle’, Trans. Ancient MonumentSoc., new series 15, 35-47.Renn, D., 1973 Norman Castles in Britain(London: John Baker).Rolland, D., 1984 ‘Le château et les châte-lains de Vic-sur-Aisne’, Fédération desSocietés d’Histoire et d’Archéologie del’Aisne 29, 139-76.Runciman, S., 1951 A History of the Cru-sades 3: the Kingdom of Acre and the laterCrusades (Cambridge University Press).



Saunders, A. D., 2002 Launceston Castle(London: English Heritage).Shoesmith, R. (ed.), 2014 Goodrich Castle:its History and Buildings (Almeley: Logas-ton).Smith, B., 2006 ‘Lacy, Hugh de, earl ofUlster (d. 1242)’, Oxford Dictionary of Na-tional Biography (Oxford University Press:accessed 30 March 2018).Smith, S. G., 2014 ‘Dolbadarn Castle, Caer-narfonshire: A Thirteenth-Century RoyalLandscape’, Archaeol. in Wales 53, 63-72.Sweetman, D., 1999 The Medieval Castlesof Ireland (Woodbridge: Boydell).Thomas, W. G., 1993 ‘The Walls of Ten-by’, Archaeol. Cambrensis 142, 1-39.Thurlby, M., 2006 Romanesque Architec-ture and Sculpture in Wales (Almeley: Lo-gaston).Tietzsch-Tyler, D., 2013 ‘King John’s Cas-tle: Staged Development, Imperfect Real-ization’, North Munster Antiq. Journ. 53,135-71.Tietzsch-Tyler, D., 2017 ‘William Mar-shal’s castle at Kilkenny in about 1395: anew reconstruction’, in J. Bradley, C. ÓDrisceoil and M. Potterton (eds), 183-200.Tietzsch-Tyler, D., 2018 ‘Innovative CastleDesign on the Western Fringe of the An-gevin Empire’, in N. Guy (ed.), 105-28.Turner, R., 2006 ‘The Upper Bailey’, in R.Turner and A. Johnson (eds), 71-80.Turner, R. and Johnson, A. (eds), 2006Chepstow Castle: its History and Buildings(Almeley: Logaston).Turner, R., Priestley, S., Coldstream, N.and Sale, B., 2006 ‘The ‘Gloriette’ in theLower Bailey’, in R. Turner and A. Johnson(eds), 135-50. Turvey, R. K., 1989 ‘Nevern

Castle: A New Interpretation’, Journ.Pembs. Historical Soc. 3, 5766.Vincent, N., 2000 ‘William Marshal, KingHenry II and the Honour of Châteauroux’,Archives: Journ. Brit. Record Assoc.25/102, 1-15.Walker, R. F., 1994 ‘The supporters ofRichard Marshal, earl of Pembroke, in therebellion of 1233-1234’, Welsh HistoryReview 17/1, 41-65.Wiles, J., 2014, “Marshal towers’ inSouth-west Wales: Innovation, Emulationand Mimicry’, Castle Studies GroupJourn. 27, 181-202.



Notes1 Christian Corvisier has however questioned

the archaeological evidence, and prefers adate in the 1150s under the influence of theCounts of Blois; other authorities have pro-posed a date in the 1160s, under the Angevinking Henry II (Corvisier 1998(2), 318-19;Corvisier 2002, 44 n. 60).

2 Those at Simiane-la-Rotonde (Alpes-de-Haute-Provence) and Ottrott (Bas-Rhin, inAlsace), lay beyond the borders of medievalFrance, within the Empire. The latter arecylindrical ‘bergfriede’ like those built inGermany from c.1160 onwards (Biller 1998,130-7; Jost 2002, 181-2; Mesqui 2013, 185;Renn 1961, 133, et al.).

3 The 1180 roll records the outlay of £13 onthe ‘repair of houses and the tower [ie. don-jon]’ at Neaufles, and £10 on the ‘repair ofthe houses, tower and gate’ at Château-sur-Epte (Stapleton 1840, 72). In the 1184 ac-count, £195 4s 8d was spent on ‘works on thetower and houses, and heightening the walland making the base of the wall’ at Neaufles,while works at Château-sur-Epte, costing£301, were confined to heightening the che-mise around the donjon and work in thebailey (Stapleton 1840, 110).

4 The internal arcade at Château-sur-Epte waspurely decorative, and never intended to car-ry vaulting (Châtelain 1983, 204). The largetwin-light window is clearly a later insertion,which interrupts the arcade while being sty-listically later, possibly early/mid thirteenth-century (Corvisier 1998(2), 145; Mesqui1993, 196; also see Châtelain 1983, 204). Itwas probably inserted under French rule,after 1204. The unusual circular oculi atNeaufles show influence from ecclesiasticalbuilding. The only other castle within whichthey have been recorded, as principal lights,is at Troyes (Aube), where they are probablymid-twelfth century (Corvisier 1998(2),490); they do however also occur in the gablewall of the eleventh-century great tower atChepstow, Mon.

5 ‘The wall in front of the tower’ is mentionedin the exchequer account of 1180 (Stapleton1840, 73); the tower is unusual in showingfour large buttresses, much smaller howeverthan the turrets employed in French donjons(see below). Henry’s ‘tower’ at Lyons-la-Forêt (Eure, Normandy), also mentioned inthe 1180 account (ibid.), may similarly havebeen a cylindrical donjon (Corvisier1998(2), 662-3).

6 When £95 was spent on ‘operacionibus tur-ris Bonavilla plancandis’ (Stapleton 1840,233). The account was rendered by Robertde Roos, lord of Helmsley Castle in York-shire and hereditary custodian of Bonneville(Stapleton 1840, 142, 233-5).

7 Corvisier and Michel de Boüard favour a dateunder King John, 1199-1204, for the muraltowers at Bonneville (Boüard 1966, 371; Cor-visier 1998(2), 62-3); Corvisier however con-siders the curtain wall itself to belong to HenryII’s reign, though it is not clear how the donjonmight have been inserted. The latter may havebeen complete by 1198 and is not mentioned inthe account roll of that year, which insteadrecords small sums spent on ‘repairing theroyal dwellings at Bonneville’ (Stapleton1844, lxxv); the roll is an isolated survival, andonly a few detached membranes survive fromJohn’s reign (see Stapleton 1844, passim).

8 Although the latter may be thirteenth-centu-ry (Corvisier 1998(2), 619-20).

9 The considerable expenditure at Pont-de-l’Arche, recorded in the mid-1190s (see be-low), represented a comprehensive rebuild.The castle, now gone, is known to haveincluded a large, free-standing round towerwhich, like the rest of the castle, cannot bedated (Launay 2015; Fig. 18); the erection ofa ‘turret’ (tourelle) was commissioned byKing Philip in 1210 (Châtelain 1991, 115-61), but the term is scarcely appropriate forthe round tower which, moreover, is free-standing within the bailey unlike the vastmajority of his post-1200 ‘Tours Philippi-ennes’ (among which it is not normally in-cluded; see Mesqui 2013, passim).

William Marshal, Pembroke Castle and Angevin design - Notes


10 Marie-Pierre Baudry notes the Angevin char-acteristics of the tower at Montreuil-Bonnin(Baudry 2002, 57-9; also see Mesqui 1997,254-5); the rib-vault was added by King Philip,by whom – alone amongst the castles of Poitou– it was held in 1206-10 (Baudry 2002, 46;Mesqui 1997, 254). The cylindrical tour-mixteat Charlieu (Loire), is suggested to have beenbuilt by Philip Augustus after his final acquisi-tion of the Auvergne from the Angevins in1210 (Deschamps 1972, 111; also see Gilling-ham 2001, 31, 89), but shows niched loops, anupper tier of large windows like Chinon’s Tourdu Moulin and, though otherwise unvaulted, abarrel-vaulted basement: could it have beenbuilt by Richard I or John?

11 Baronial Courville-sur-Eure and Montland-on (Eure-et-Loir), though under French over-lordship, may show Angevin influence intheir later twelfth-century donjons (Corvisier1998(2), 276-8, 475-8). Angevin influencemay also be apparent in the free-standing,circular ‘tours-beffrois’ of central Aquitaine(Mesqui 2013, 103), which appear to havebeen built during the late twelfth and earlythirteenth centuries.

12 Perhaps surprisingly, King Stephen of Blois(1135-54) is not known to have built one ineither Britain or France, and the cylindricaltower at New Buckenham (Norfolk), built byhis follower William d’Albini in 1138-46, isthought to be purely local in context (Hislop2016, 113-14, et al.). A cylindrical donjonhas been suggested at Sauvey Castle (Leics.),which was begun by King John around 1211,but the site is little-understood (Colvin 1963,829; King 1983, 255; Page 1907, 249-50;Renn 1961, 131).

13 Towers of this kind can be distinguished fromearlier rectangular donjons such as Portchester(Hants.), which were built before the principleof flanking was fully appreciated and are onlyincidentally peripheral and engaged.

14 The Record Tower’s definition as a donjonhas been challenged by, among others, Har-old Leask and Con Manning (Manning 2003,72; also see Murtagh 2017, 162, 175). Dan

Tietzsch-Tyler regards the ‘dominant tower’,in general, as having ‘taken the place of the . .. keep’ as ‘essentially a public space’ (Tietz-sch-Tyler 2017, 186).

15 The date of the Barnard chamber-tower is stilla matter of debate. In a recent paper, MalcolmHislop generally favoured David Austin’sdate in the 1190s. However, he noted thatfinancial circumstances may instead argue adate-range of c. 1208-28 (Hislop 2018, 99) –which is compatible with the stylistic evi-dence and preferred here.

16 Affinities with the buttressed cathedrals ofLanguedoc have also been suggested, eg. Albiin Tarn and Agde in Hérault (Mesqui 1993,329-30), though Richard I is not recorded theregion (see Landon 1935).

17 And Richard I may have built a polygonaldonjon at Radepont (Eure), in Normandy(Corvisier 1998(2), 133 n. 38).

18 The curtain wall inturns from its oval trace toconnect with the tower, which sits well backfrom the edge of the enclosure.

19 As it is relative to the buttresses added to thetower at Gisors, by Henry II, in c.1170-1184(Mesqui 2013, 199; Stapleton 1840, 72, 110).Conisbrough and Mortemer probably influencedsimilarly buttressed donjons in Normandy andMaine, at baronial Condé-sur-Noireau (Calva-dos), Le Plessis-Lastelle (Manche) and Saint-Rémy-du-Plain, Sarthe (Corvisier 1998(2), 256-62, 507-10, 581-5).

20 The tower at Gournay-sur-Marne (Seine-Saint-Denis, Île-de-France) seems also to have been apure cylinder; it is suggested to have been royal,and earlier twelfth-century (Corviser 1998(2),352-5), but has now disappeared.

21 An isolated circular tower, of unknown form,was also built at the royal palace in Paris(Mesqui 2011, 311 et al.).

22 As in Angevin tours-mixtes, however, multi-ple functions are implicit in their design and itis possible that some saw royal use on an adhoc basis (Jean Mesqui, pers. comm.); thetower at Vernon (the ‘Tour des Archives’), for



example, may have housed King Philip’streasury during the late 1190s (see Powicke161-2 and n., 401-2, 430).

23 John was in Paris, with Philip, in early 1193(Holden et al. 2006, 126-7; Landon 1935, 723;Powicke 1913, 143-4). He again visited Parisduring the Le Goulet peace in July 1201 (Church2015, 97), but this was his last time in the Frenchroyal domain, and many aspects of Philippienneplanning had yet to properly emerge. Marshalwas not present on either occasion.

24 While most authorities agree that rebuildingin masonry commenced at Kilkenny in 1207,it is possible that work had already beganunder Marshal’s seneschal Geoffrey FitzRob-ert (see Crouch 2016, 120 and n. 34): thecastle was strong enough to withstand siegein early 1208. But while its crenellations(‘kerneals’) are mentioned in a description ofthe attack (Holden et al. 2004, 193) – sug-gesting it was more-or-less complete – theaccount was written nearly twenty years later,and therefore unreliable.

25 Nevertheless it has been suggested that itmay, at first, have been isolated (Tietzsch-Tyler 2018, 126 n. 21).

26 Con Manning felt that the circular southwest‘Bermingham’ tower was the first element to bebuilt at Dublin (Manning 2003, 90), and that theRecord Tower may be rather later; ‘constructionat Dublin mainly spanned the period 1213-c.1230’ (Manning 2003, 72). Tadhg O’Keeffehas suggested that instead it was the RecordTower that was begun soon after 1204, and thatit influenced Kilkenny (O’Keeffe 2009, 283-4),but keeps an open mind.

27 An attribute shared by the later dominanttowers at Cilgerran, Laugharne, Manorbierand Narberth in west Wales (see below). Withthe exception of Narberth, these towers – likethe Pembroke donjon – also lack latrines(Wiles 2014, 190, 192).

28 John Goodall has for example observed thatno Tour Philippienne, including the Louvre,can have influenced Pembroke’s donjon(Goodall 2011, 162).

29 Marshal may have spent time in the Capetianlands with Henry the Young King in the early1180s, while he attended two tournaments nearCompiègne in winter 1182-3 (Holden et al.2002, 257-81, 305), when he may conceivablyhave seen the pre-Philippienne donjon there – atleast from the outside. He had, in 1188, beenpromised the hand of the heiress to Châteaur-oux, in Berry, for which Henry II negotiatedwith the canons of Bourges (Vincent 2000, 5-8),where the Tour Philippienne was under way;there is however no evidence that Marshal him-self accompanied the king. But it is possible thathe saw Philip’s tower at Gisors being built whilecampaigning there in September 1198 (Holdenet al. 2004, 47-55), though Richard I’s forcesare not recorded near Vernon.

30 His absence meant he was unable to influencehis Normandy lordships, and while it is clearthat his followers maintained his interests, hewas reliant on ‘the good will (or otherwise) ofCapetian officers and his local dependants’;contact was minimal, and there was no returnto the cross-channel lifestyle Marshal hadenjoyed before 1204 (Power 2003, 211-13,216, 224).

31 Simple loops were employed in PlantagenetGascony until the end of the thirteenth centu-ry, for instance at Sauveterre-la-Lémance(Lot-et-Garonne), built by the English Crownduring the 1280s (Mesqui 2013, 69 and fig.).

32 The order to strengthen Arques was given inMarch 1201, to Jordan de Sauqueville whowas Marshal’s tenant in Normandy and one ofhis leading mesnie knights (Hardy 1835b, 35;Power 2003, 219, 224). Some of this workwill, moreover, have been undertaken whenthe castle was in Marshal’s direct custodyApril-December 1202.

33 Marshal was possibly with King John duringfurther stays at Bonneville in October 1201and March 1203 (Hardy 1835a, 2, 27).

34 The isolated Norman exchequer roll of 1198records £20 spent on ‘repairing the dwellingsof the castle and the bridge’ at Domfront(Stapleton 1844, lix), which may have been



the start of a wider programme on the castlegate; no expenditure had been recorded in theroll of 1195 (Stapleton 1840). Work may havebeen more-or-less complete by summer 1202when £100 was spent on the ‘towers andhourding’, though unspecified expenditure atthe castle continued through 1203 (Hardy1835b, 50, 79, 84).

35 Triangular loops were also employed in thegatehouse at Carrickfergus Castle, probablybuilt by King John (see below).

36 They similarly appear in baronial work from1199-1214 at Warkworth (Goodall 2015, 26),and elsewhere.

37 It is quite possible that the ‘Hanbury Tower’ atCaerleon, within which these loops occur, is thework of the younger William Marshal, who heldCaerleon 1219-23, and again 1226-36 (Holdenet al. 2006, 187). It may belong to the samecampaign as the structure on the motte, which isnormally considered to have been a shell-keep.William Marshal II is thought to have built theshell-keep at Wiston, Pembs., soon after 1220,and may have been responsible for the one atCarmarthen Castle during his tenure there 1223-26 (discussed in Ludlow 2014, 18-23). AYounger Marshal liking for the shell-keep –which is not otherwise a feature of west Wales– may be confirmed by its possible use at StClears Castle, Carms., held by the Marshalsfrom 1230 until 1245 (Pat. Rolls 1226-32, 339).

38 Offsets are absent from Philippienne work at,for example, Beauvais, Dourdan, Gisors, La-on, Lillebonne, Montdidier, Péronne and Vil-leneuve-sur-Yonne.

39 Neither does any Marshal work show thestring-courses that are depicted, in a late-me-dieval illustration, throughout the Louvre (Fl-eury and Kruta 2000, 68).

40 The form was also adopted by Hubert deBurgh, who had led the defence of AngevinPoitou in 1202-4 and was later seneschal ofthe province, in the inner ward towers atWhite Castle (Mon.), built c. 1229-39 (seeRemfry 2011, 221-2).

41 They are more frequent in French baronialwork of the thirteenth century e.g. at La Fère-en-Tardenois (Aisne), in Picardy, begun c.1206(Mesqui 1997, 168 and fig.), Saint-Vérain(Nièvre), where the slightly later towers show‘Angevin’-style niched embrasures (Corvisier1998(2), 594-5), and in Coucy’s outer defenc-es, 1220-30s (Mesqui 1997, 135-8). Their usein the outer ward at Harcourt (Eure), soon after1204, is thought to reflect continued Angevininfluence (Mesqui 1997, 457).

42 Its employment at Pembroke, by William deValence, may have been a conscious referenceto his marriage bonds and Marshal inheri-tance, in the heartland of the Marshal buildingtradition where it would be fully understood.

43 Richard’s itinerary does not, however, allowfor a stay at Druyes; he was at Donzy on 1 Julyand had reached Vézelay, over 30 miles away,by 2 July (Landon 1935, 36).

44 Marshal was not with them: he remained inBritain as one of four co-justiciars during Rich-ard’s absence (Crouch 2016, 90-4).

45 They include Deir al-Balah (‘Darum’), Tel-es-Safi (‘Blanchegard’), Yazur (‘Casel desPlains’ and ‘Casel de Maen’), Yibna (‘Ibelin’)and Ashkelon itself (Landon 1935, 57-62).

46 As one of only two surviving rolls for the1190s, it may not record the total expenditurethere, while no particulars are given.

47 Complicating the issue is the statement, by thechronicler William the Breton, that the castlewas slighted by its Norman defenders inspring 1204; it is far far from certain howeverthat any damage was extensive (Powicke1913, 373).

48 Recorded expenditure on Richard’s easternNormandy castles, which is doubtless incom-plete, amounted to around £21,500; an addi-tional £46,000 was spent on the combinedworks at Les Andelys, and over £5000 onfortifications at Eu, Seine-Maritime (Stapleton1840 and 1844, passim). Major work by Rich-ard I has also been suggested in Poitou(Baudry 2002, 45, 67-9; see above).



49 Nevertheless, whilst there Marshal had com-mitted his body to the Templars (Holden et al.2004, 415), meaning that he may have spenttime with the order.

50 The influence of existing Roman fortificationsmust be taken into account (Knight 2018, 157;Mesqui 2013, 37, et al.); it is manifest atPortchester Castle (Hants.) where the rectangu-lar outline of the early twelfth-century innerward was clearly dictated by the Roman fortwithin which it was built. The use of symmet-rical, quadrangular planning in monastic build-ing may also be significant: it was probably themodel for Bishop Roger of Salisbury’s court-yard plans at Old Sarum and Sherborne OldCastle, both from the 1120s-30s.

51 King John stayed at Niort on various datesfrom April through to September 1214, whenhe is also recorded at nearby Parthenay (Har-dy 1833, 166, 168, 170-2; Hardy 1835a, 113-14, 119, 122, 139). Marshal remained in Brit-ain during both 1206 and 1214, though hisson Richard was possibly with the king dur-ing the latter campaign (Crouch 2015, 11;Holden et al. 2004, 237).

52 Baileys with varying degrees of rectilinearityare seen in other earthwork castles from thelate eleventh and twelfth centuries, for in-stance Carmarthen, Lincoln, St Clears(Carms.), Warkworth and Warwick (Ludlow2014, 178).

53 Shoesmith’s plan shows this early tower withthe same spur-buttresses as the latethirteenth/fourteenth century towers (Shoe-smith 2014, 130 fig.), for which howeverthere appears to be no excavated evidence.

54 Marshal may have seen geometric castles inFrance during his 1216 visit although, arguably,he is most likely to have met King Philip atPont-de-l’Arche (see above), an appropriately‘borderland’ location which was also the venuefor his widow’s meeting with Philip in July 1219(Crouch 2015, 190-3; Holden et al. 2006, 195).

55 The elder Marshal did not receive Dunamaseuntil 1208 (Hardy 1835a, 80; Holden et al.

2006, 157); two years later, it was confiscatedby King John (Holden et al. 2004, 217-21)and was not returned until 1216 (Hardy 1835a,180, 184). It is therefore unlikely that Marshalwas responsible for any building there: thehall, and possibly the inner curtain, were al-ready in existence, while his priorities layelsewhere 1216-19 (Hodkinson 2003, 47-8;Holden et al. 2006, 157). Brian Hodkinsonbelieves the inner barbican and gatehouse –which was equipped with a drawbridge – wereinstead built between 1219 and 1231 (Hodkin-son 2003, 48).

56 Also distinctly Angevin is the system oftunnels beneath the defences at Arques, ex-tending into the castle ditch (Langeuin 2002,364-5), cf. Dover, Nottingham etc. (Baudry2002, 61; Hislop 2016, 19, 155).

57 Their position astride the curtain is analo-gous with ‘Marshal’ towers (see above) but,like the similarity of these gate-towers to thedouble-entried Tours Philippiennes, is prob-ably incidental (see Mesqui 2013, 318).

58 It is possible that grooves were present but,as in Warkworth’s gatehouse of 1199-1214,did not descend to ground level and so notpreserved.

59 Cf. the similar gate-tower in the Armeniancastle at Korykos, Turkey, built c.1200 (De-nys Pringle, pers. comm.).

60 These ‘horseshoe’ gate-towers can be distin-guished from semicircular barbicans, whichwere open enclosures and perhaps primarilyadministrative, for holding courts etc. (seeDay and Ludlow 2016, 94). The earliest inBritain appears to be at the Tower of Lon-don, from 1275-81 (Impey and Parnell 2011,34-6), followed at Goodrich and Pembroke’souter gate (c.1290-1320). A similar barbicanhad been built by the mid-thirteenth centuryat Carcassonne, Aude (Mesqui 2013, 354fig.).

61 Neil Guy also notes their use by EmperorFrederick II in early thirteenth-centurysouthern Italy (Neil Guy, pers. comm.).



62 Pivot-stones have yet to be reported fromother contemporary work in the British Isles.However, they were clearly vulnerable torobbing and loss; empty sockets in the gate-way at Manorbier (Pembs.), built by a Mar-shal tenant c.1230, may have held somethingsimilar (Neil Guy, pers. comm.). Marshalinfluence presumably lies behind their adop-tion in Ireland, which persisted in tower-houses into the seventeenth century (Leask1964, 97-8).

63 The solid-fronted, semicircular towers in themain gatehouse to Montgomery’s inner wardrecurve sharply at their junction with both thecurtain and entrance arch, prompting com-parisons with Marshal’s Chepstow gatehouse(Hislop 2016, 167-8); a similar, if less-pro-nounced shape was again employed by deBurgh at White Castle, c. 1229-32 (see Rem-fry 2011, 219-21). However, we saw abovethat this shape had been used by the Angevinkings, as were solid rounded gate-towers; theChepstow gate-towers, moreover, do not re-curve at the entry. De Burgh held Montgom-ery for the Crown. Its gate-towers areelongated as at King John’s Kenilworth andin the slightly earlier gate at royal Pevensey(see below), while nearly three-quarter roundgate-towers were used by the Crown in thelate thirteenth century, at Harlech and Rhud-dlan.

64 This lasting influence is confirmed by thelarge window in Caldicot’s southeast tower,which is a close copy of the late thirteenth-century windows in Roger Bigod’s ‘Glori-ette’ at Chepstow.

65 An earlier date in the 1240s has howeverbeen suggested (Goodall 2011, 183, 201), ie.not long after Caldicot.

66 Hislop points out that Barnard Castle was held,between 1278 and 1296, by John de Warenne’sson-in-law John Balliol the Younger.

67 It may be significant that Marshal was atChepstow in July 1217 (Hardy 1833, 314;Pat. Rolls 1216-25, 79), and September 1218(Hardy 1833, 370).

68 But it, too, may post-1219 (Tietzsch-Tyler2018, 127).

69 The remains were sketched and photographedby the author in 2016 and 2017. Thanks toDan Tietzsch-Tyler and Jean Mesqui for dis-cussing the results with me.

70 He had acquired Longueville through mar-riage in 1189 (Crouch 2015, 465-6). The Mar-shals were again resident lords in Normandybetween 1220 and 1231, through William theelder’s son Richard (see below), and whilesolid-towered gatehouses were still under con-struction at this time eg. at Launceston (Corn-wall), Longtown (Herefs.), Montgomery andelsewhere, they generally show portcullises(Knight 1993, 127-8; Saunders 2002, 6).

71 The gatehouse at baronial Bricquebec, from theend of the twelfth century (see above), showssimilar towers but with open backs.

72 The baronial twin-towered gatehouses atHelmsley and Warkworth are from the samegeneral period but cannot be closely dated.See Neil Guy’s recent paper for a well-bal-anced review of contemporary gatehouse de-velopment in England and Wales (Guy2016a).

73 King John formally thanked Marshal for hisintervention with the Pope in summer 1212(Hardy 1833, 132; Sweetman 1875, 72-4), butwas reluctant to allow his return, commandinghim to remain in Ireland while warning crownofficers to be vigilant against any harm that‘may be worked against you from the lands orjurisdiction of Earl William Marshal’ (Hardy1833, 121-2; Sweetman 1875, 72-3). Mar-shal’s hostage sons were not released by theking until late 1212 or early 1213 (Holden etal. 2006, 159).

74 Guy de Cultura, a mesnie knight of WilliamMarshal, in Britain after 1204 (Crouch 2015,80-1, 108-11, 182-3), was from Les Andelys(Cultura = Petit Andely; see Pitte 2002, 167-9).

75 The witnesses narrow the date to 1203-10, andinclude Nicholas Avenel, Marshal’s stewardof Pembroke in 1202-3 (Brewer 1863, 227;



Davies 1946, 322, 327) – probably still inoffice in 1204 – and Robert Fitz-Richard, lordof Haverfordwest until 1210. Marshal is nototherwise known to have visited west Walesduring these years.

76 A loose fragment of dogtooth is also re-usedat Usk Castle (Knight 1977, 147), and proba-bly also Marshal-period.

77 Dogtooth seems nevertheless to have its ori-gins in northern France where it was em-ployed from the 1130s until the end of thetwelfth century, and could formerly be seen atThérouanne Cathedral, Nord (Hourihane2012, 300), demolished in the 1550s.

78 The West Country was the main architecturalinfluence in twelfth-century Wales and, viathe Marcher lords, Bristol and the Welshports, also gained considerable agency in Ire-land (Brakspear 1931, 4; Thurlby 2006, pas-sim). Marshal’s closest ties were moreoverwith this region, which represented his mainpower base and the source of many of hishousehold knights and followers (Crouch2016, 217-18, 231-4); some of these, for in-stance the Berkeleys of Gloucestershire, builtin a notable West Country style in their ownhouses and castles eg. the Great Hall at Berke-ley Castle, from c.1180 (Emery 2006, 58).

79 Marshal’s grant to Monkton Priory, of the tithesof three mills (Caley et al. 1846, 320-1; Crouch2015, 161-2), may have contributed to this work.

80 The tower at Charlieu (Loire), which may beAngevin, also has a barrel-vaulted basement(see Table 1), while the Tour du Moulin fea-tures a ‘hybrid’ barrel/rib-vault as used byRichard I in the Bell Tower at London. Theentrance-floor vault at Châtillon-sur-Indre iseither primary, under influence from Blois(Corvisier 1998(2), 197, 201), or a secondaryinsertion when the donjon was heightened inthe thirteenth century (Deyres 1984, 366, 376-7). The ground-floor vault in the cylindricaldonjon at Inchiquin, Co. Cork, also appears tobe secondary (McNeill 1997, 241), as is theribbed vault at Conches-en-Ouche (Corvisier1998(2), 255).

81 Additional saucer-vaults have however beensuggested at the Barnard chamber-tower,based on antiquarian evidence from the six-teenth and early nineteenth centuries (Hislop2018, 93 and n. 6, 96, 102).

82 Similar vaulting was used in the polygonaldonjon at Châtillon-Coligny (Loiret, Île-de-France), in 1180-90, inspired from Blois(Mesqui 2003, 111 fig.).

83 It may be noted moreover that the alternatingarrangement of stairways at Conisbroughseems designed to place the maximum socialrestriction on access to the upper floors (Mc-Neill 2006, 123).

84 It has been suggested that the donjon at Or-ford Castle, Suffolk, carried a conical wood-en ceiling or vault as early as the 1170s,perhaps supported on a series of arcade posts(Heslop 2003, 292-5).

85 Derek Renn noted the general similarity be-tween the three (Renn 1968, 41-3).

86 Marshal’s eldest daughter Matilda marriedHamelin’s son William, but not until 1225(Crouch 2015, 39). Marshal’s only attestedappearance north of Nottingham was inYork, in February 1206 (Hardy 1837, 162).

87 Also absent from Château Gaillard and Pem-broke are the joggled lintels seen at Conis-brough; the same technique was used byHenry II in his donjons at Dover, Orford andTickhill, Yorks. (Brindle and Sadraei 2015,8), demonstrating Conisbrough’s close rela-tionship with works of the Crown.

88 Rib-vaults were infrequent in English roundtowers until the mid-thirteenth century, andwere never widely used in donjons or tours-mixtes, being largely confined to mural tow-ers – and those of D-shaped, rather thancircular plan (see Ludlow 2018, 254 and n.11).

89 The vaults in the late twelfth-century hall-block at Manorbier Castle, Pembs., are sec-ondary insertions (King and Perks 1970, 95-6, 116).



90 In addition, several late-medieval church tow-ers in south Pembrokeshire show ‘domed’summit vaults, eg. St Petrox and St Twynnells(Lloyd et al. 2004, 446-7), while a similartower dome in Tenby’s town walls belongs tothe late fifteenth century (Thomas 1993, 13).

91 The Marshal/west Wales tradition, referred tothroughout, was probably not a formal schoolwith dedicated lodges, but may have been alooser network of professional links throughtime, as well as space, and lasting over 200years. It was perhaps largely driven by hom-age and emulation (see Wiles 2014).

92 A date between 1216 and 1231 has howeveralso been suggested (Guy 2017, 29).

93 Semicircular turrets, sometimes housingstairs, were also present in the French don-jons at La Queue-en-Brie (Val-de-Marne),and Montchauvet and Montfort-l’Amaury(Yvelines); they belong to a group of unusu-al, non-rectilinear towers built by the Mont-forts during the twelfth century (Corvisier1998(2), 442-3, 461-3, 541-50, 580).

94 Ardfinnan was periodically held of the Crownas a tenant lordship, but was in John’s ownhands by 1210 (McNeill 1997, 31, 233). Thetower shows cruciform loops, as employed bythe Angevins by the mid-1190s (see above), inembrasures with slightly pointed heads.

95 It has been suggested that the simple chamferat Lyonshall may indicate an even earlier date,in the 1180s (see Pamela Marshall’s com-ments in Guy 2017, 121). However, the cham-fered string-course was a very conservativeand persistent form seen, inter alia, in thebasal string-course of the dominant tower atNarberth Castle, Pembs., which was almostcertainly built by Roger Mortimer of Wig-more – using masons from the Marches – after1257 (Ludlow 2003, 13, 24). It was widelyused in Wales, particularly in church towers,into the sixteenth century.

96 Skenfrith’s conservative round-headedopenings may be a deliberate reference toAngevin prototypes, like so much work ofde Burgh’s work including White Castle and

in the Constable’s Gate at Dover. This dualinfluence may also lie behind the similardonjon on the motte at Chartley (Staffs.),which was probably built in the 1220s byRanulf, earl of Chester (Goodall 2011, 180-1; Hislop 2016, 125), and also shows a semi-circular turret. The earl had spent most of1199–1204 in Normandy where he pos-sessed extensive estates, while his constableof Chester had defended Château Gaillard(Powicke 1913, 365, 379-80, 446, 491).Chartley’s overall plan-form has been com-pared with King Philip II’s castle at Montl-héry, Essone (as noted by Duffy 2017, 312and fig., 314), but any similarity is slight andincidental: Montlhéry has no motte, andshows a Philippienne symmetry entirely ab-sent from Chartley where the towers andgatehouse are also of very different form.

97 The small, isolated round tower at TenbyCastle, which apparently has a summit-vault(King 1977, 164), with a ground-floor entryand rectangular stair-annexe that may besecondary, is possibly analogous to Nevern,but in the absence of diagnostic detail cannotbe closely dated. While it conceivably be-longs to the earlier thirteenth century, mostauthorities have followed the Royal Com-mission’s preference for a date in the mid-late thirteenth century (eg. Goodall 2011,208). The possible small, half-timberedround tower on the motte at Carmarthen canbe neither fully-interpreted nor dated (Lud-low 2014, 179-81).

98 King Philip is attested at a number of loca-tions during the period of the elder Mar-shal’s visit, including Melun, which is 25kmnorth of Nemours (Samaran and Nortier1979, 27-32). However, the meeting is morelikely to have occurred at Compiègne orPont-de-l’Arche (see above).

99 Richard Marshal may also have been withPhilip in summer 1220, between receiving andtaking formal possession of his Longuevillelordship, when the king again stayed at Nem-ours (Samaran and Nortier 1979, 304-5).



100 The sources suggest that Richard Marshalmade attempts, during the Anglo-Frenchdétente of 1219-24, to revive his father’scross-channel lifestyle (Hardy 1844a, 98-9;see Power 2003, 210-11). He had beengranted the manor of Long Crendon(Bucks.), by his elder brother, in 1219-22(Power 2003, 210), while a number of man-dates were issued in 1225-6, after the re-sumption of hostilities, ordering his shipsfrom Normandy to be impounded (Hardy1844a, 47-8, 149).

101 The towered keep at Terryglass, Co. Tipper-ary, had been built by 1333 (McNeill 1997,118-19). The other four are either lost (Wex-ford, which had been built by 1323), latemedieval (Dunmoe, Co. Meath and Enniscor-thy, Co. Wexford) or undated (Delvin, Co.Westmeath); see Colfer 2013, 52; Sweetman1999, 39, 60; McNeill 1997, 118. O’Keeffeand Coughlan regarded Ferns as belonging tothe earlier part of William de Valence’s ten-ure, during the mid-thirteenth century(O’Keeffe and Coughlan 2003, 147). Howev-er, they allow that certain of its features – thecross-loops and trilobe-headed twin-lightwindows – would normally indicate a laterdate. To these might be added the large circu-lar oillets (cf. Valence’s Goodrich); thecolour-banded masonry (cf. Caernarfon andHaverfordwest); and the similarity of thewindows to those in Roger Bigod’s Glorietteat Chepstow, from the early 1280s (Turner etal. 2006, 135-7), and in Queen Eleanor’sGloriette at Leeds Castle, Kent, from 1278-90 (Goodall 2011, 236-7 and fig.). Filletedcorbel-shafts, as in the Ferns chapel, werestill being used into the 1320s in the great hallat Caerphilly, while their facetted springersclosely resemble those in the cellar of Chep-stow’s Gloriette. Ferns may then belong to alater period of Valence’s tenure, perhaps the1280s-early 1290s.

102 There are further similarities between Nem-ours and Irish examples, including the use ofone corner turret as a chapel at Nemours(Mesqui 2013, 128, 207), Ferns (O’Keeffe

and Coughlan 2003, 138-43) and probablyLea (Dempsey 2017, 242). No central pierhas been identified at Lea, Carlow or Ferns,unlike Nemours where, however, it may besecondary (Mesqui 2013, 128, 214).

103 The castle is not named. In 1230-31, WilliamMarshal II was quit of 80 days’ service, thistime towards work on the castle at Castle-comer, Co. Kilkenny (Sweetman 1875, 270,278).

104 Richard himself was probably the source forthe poem’s glowing testimonials to his fa-ther, made by French knights after the lat-ter’s death in 1219 (Holden et al. 2004, 469;Holden et al. 2006, 25, 195).

105 David King, while noting that the two arebonded higher up (King 1978, 118), does notmention that the curtain was heightened,where it joins the tower, when the latter wasadded (see Ludlow 1990, 28 and fig.; Fig. 35).

106 Also of interest is Richard Marshal’s alliancewith Llywelyn in 1233-4 (Neil Guy, pers.comm. See Walker 1994, 43-4, 62-3;).

107 The Dolbadarn sockets have an alternativeinterpretation as supports for a timber stairleading to an appearance balcony (Smith2014, 68-9); an appearance doorway is sug-gested in Pembroke’s donjon (see Appendix1), but by itself this does not imply anyMarshal associations at Dolbadarn. A rela-tionship between Dolbadarn and the slightlyearlier donjon at Clogh Oughter has beensuggested by Con Manning, via the marriageof William Gorm de Lacy, of Clogh Ough-ter, to Llywelyn ap Iorwerth’s daughterGwenllian (Manning 2013, 198-9).



Table 1 – Circular and polygonal great towers in France before c. 1203 (compiledfrom Corvisier 1998, Baudry 2002, Mesqui 2013 and others) 1/3

Ang. - Angevin; FR. - French; Vest. - Vestigial; * - Tour Philippienne.

William Marshal, Pembroke Castle and Angevin design - Table 1


Table 1 – Circular and polygonal great towers in France before c. 1203 (compiled fromCorvisier 1998, Baudry 2002, Mesqui 2013 and others), continued, 2/3



Table 1 – Circular and polygonal great towers in France before c. 1203 (compiled fromCorvisier 1998, Baudry 2002, Mesqui 2013 and others) continued 3/3



APPENDIX 1:Pembroke’s donjon: form and function

The circumstances behind the erection ofPembroke’s donjon, discussed above, sug-gest that it was primarily celebratory andcommemorative like the donjons at e.g.Coucy (Aisne, in Picardy), Hedingham (Es-sex), Rochester (Kent), Castle Rising (Nor-folk) and Dundrum (Co. Down), all ofwhich were built to mark major acquisi-tions, social advance and upturns in status –frequently, like Marshal, through marriage(Dixon 1998, 55-6; Marshall 2002a, 146;McNeill 2003, 96, 98). Though fireplaces are present at first- andsecond-floor level, there appears to have beenno provision for a latrine, indicating that –like the great towers at e.g. Château Gaillard,Chepstow, Hedingham (Essex) and the Tow-er of London – it was not intended for resi-dential use (see Dixon 2002, 11; Marshall2002b, 32; Marshall 2016, 171) and, in this,it may be like many circular donjons. Thethird, uppermost floor is unheated and showsno arrangements consistent with use as abedchamber. Neither of the two fireplacesserved a food-reheating area, unlike the don-jons at e.g. Norwich, Castle Rising (Norfolk),Bowes (Yorks.) and Sainte-Suzanne in May-enne (Marshall 2002a, 149; Marshall 2002b,30), and many Tours Philippiennes includingDourdan and Lillebonne (Mesqui 2013, 163);neither is there is any evidence of a well. Allthese attributes, or lack of them, indicatehighly restricted and mediated use, probablylimited to ceremonial occasions (see Mar-shall 2002a; Marshall 2002b, 30). A residen-tial chamber-block for the earl alreadyexisted at Pembroke, in the form of the Nor-man Hall (Fig. 16); Marshal’s communal halland steward’s accommodation appear to havebeen rebuilt and repurposed in the late thir-teenth century (Day and Ludlow 2016, 66-8).

Situated for maximum effect on thehighest part of the headland, and immedi-ately behind the defences, Pembroke’sdonjon was also a high-visibility architec-tural reminder of lordship, by an earl whowas normally absentee (see Marshall2002a, 142; Marshall 2002b, 35; Marshall2016, 161). The donjon may not, in fact,have seen any use under William Marshalwho was in Ireland 1207-13 and notknown to have visited west Walesthereafter.1 Several visits by the youngerMarshals are however known or surmised,when the donjon may have fulfilled itsintended role. Nevertheless, even whenthe lord was in residence, it is suggestedthat many larger donjons were used infre-quently – possibly limited, in some cases,to a single, specific occasion (Marshall2016, 160).2 But, at the same time, PhilipDixon’s reminder that designed functionmay have differed from actual use must beborn in mind (Dixon 2002, 11). At second-floor level is a narrow externaldoorway, beneath which are two largejoist-sockets in the external face, presum-ably to support timber decking. A jettied,timber latrine, as suggested in e.g. thedonjon at Caldicot, Mon. (Morgan andWakeman 1854, 22), is unlikely – theentrance passage is straight, without thedog-leg normally seen in latrine passagesincluding Caldicot (see Wood 1983, 378).Nor is it likely that the doorway led onto atimber bridge to the inner curtain parapet,providing access to the latrine turret at thenorth end of the curtain.3 The curtain issecondary to the donjon (though probablycommenced during its construction) and,as originally built, its parapet lay fourmetres below the level of the doorwaythreshold (Fig. 35).4 The entrance passagelies at an awkward, oblique angle to the

William Marshal, Pembroke Castle and Angevin design - Appendix 1



Fig. 36 – Floor plans of the donjon at Pembroke (radial joist arrangement after Rick Turner).


curtain (Fig. 36), and the necessary combi-nation of stairway and bridge would havebeen a complex construction that is notevident in the surviving fabric. The turret,moreover, lies nearly 35 metres away fromthe doorway (Fig. 16), while the latrineswithin could not be accessed from parapetlevel; a possible latrine in the HorseshoeGate is similarly distant. There is no roof-crease above the doorway, showing that thetimber superstructure was open to theelements,5 while the entry itself is not fur-nished with a drawbar-socket, arguingagainst communication with any other partof the castle. It might moreover be askedwhy, if latrines were required, they werenot built into the wall-thickness as in thepre-existing donjons at e.g. Orford andScarborough – where they discharged intothe inner ward rather than outside the castle– or latrines with intramural cess-pits (‘pit-latrines’), as employed in the donjons ate.g. Neaufles, Coucy and Dover (Corvisier1998(2), 490; Mesqui 2013, 169-73; Brin-dle 2015, 15-21), and perhaps at Marshal’sPembroke itself, in the junction betweenthe Horseshoe gatetower and the curtain(Fig. 16).6 Finally, the curtain here was,until the mid-thirteenth century, the outerdefence of the castle lying alongside thetown – a very exposed, public location forsuch a private function. Its very visibility may however provide aclue to the function of the entry: it is mostlikely to have led to an appearance balcony,facing the main street of the town and theprobable site of the early marketplace, fromwhich the earl could be seen above thecurtain parapet. It is very tall, and rathernarrow – consistent with this role – and hasan oolite surround of high quality. The ap-pearance doorway and balcony were a long-established tradition, with Carolingian ori-

gins (Marshall 2016, 165). It is evident inthe early rectangular donjons of Anjou(ibid.; Gregory and Liddiard 2016, 149),and in the cylindrical donjons at Frétevaland Châtillon-sur-Indre (Corvisier 1998(2)199, 201; Mesqui 2013, 129, 131 fig.),which Marshal may have seen in the1180s-90s (see above). It was also seen inBritain during the twelfth century, wherethe balcony at Norwich, as at Pembroke,overlooked the marketplace (Gregory andLiddiard 2016, 149). It may be significantthat an appearance doorway already exist-ed at Marshal’s Chepstow, in the earliergreat tower (Turner, Jones-Jenkins andPriestley 2006, 39-40). At Pembroke, a single spiral stair connectsall four storeys and the parapet; no separateline of communication existed between lev-els, according to social rank, suggesting thataccess to the entire donjon was restricted tothose above a certain status (see McNeill2006, 123, 125). In addition, the first-floor(entry-level) chamber had to be entered toreach the stair, suggesting the two upper-most floors – whose entries were draw-barred from inside, against the stair – wereout-of-bounds. The donjon therefore ap-pears to represent a space within whichaccess was controlled and mediated, to en-ter which was a mark of great privilege (seeMarshall 2002b, 33). The eight narrow-splayed openings inthe body of the donjon are frequentlyreferred to as loops. The embrasures how-ever are four metres long, but most areless than a metre wide to the interior andaverage only 1.5m in height (some aremuch lower), while their sills lie up to ametre above floor level. No functionaltraverse is possible within them, whilethey clearly could not be used by an arch-er standing within the interior of the don-



jon itself. And while the sills showplunges, they are very shallow: the groundwithin a 25-metre radius could be neitherseen nor controlled. Jean Mesqui notesthat, in general, donjons ‘do not include . .. arrowloops; when there are external slots,they are for lighting’, consistent with theirrestricted access as stressed by Philip Dix-on (Dixon 2002, 10; Mesqui 1993, 252).So the Pembroke openings are more likelyto be slit-lights or possibly, as Derek Rennsuggested, for ventilation (Renn 1968, 39).They are, nonetheless, an unusual featureabove basement level in free-standing cir-cular donjons, the closest comparisons be-ing again in Angevin France e.g. LeBrandon (Indre-et-Loire in Touraine), LeChâtelier and Neaufles, and at Caldicot(Mon.) and Dundrum (Co. Down); at Pem-broke, their design may have been chosenas a deliberate foil to the two windows, thevisual impact of which is enhanced by thesimplicity of the lights.The interiorThe ground floor forms a basement that,originally, was entered solely from the firstfloor via the spiral stair. As in the donjonsat e.g. Dinefwr and Dryslwyn (Carms.),Dundrum (Co. Down) and Skenfrith(Mon.), the present ground-floor entry issecondary (King 1978, 99-100; Knight2009, 30; Guy 2015, 47-9; Renn 1968, 38;Rees and Caple 2007, 30, 51); possiblylate-medieval or sixteenth-century, it is anindication of the changed use and status ofthe donjon also apparent in other altera-tions discussed below. The chamber is un-lit, and otherwise featureless. Pembroke’slimited use by Marshal may argue againstit housing a treasury and/or archive store,as in the donjons at Rouen and elsewhere(Holden et al. 2004, 89; Renn 1968, 37),while the sources suggest that their main

treasury in Wales and the west was atChepstow (Crouch 2006, 49; Holden et al.2006, 36). Nor was it particularly secure.It is possible that, in addition to its prima-ry function of adding height to the donjon,it merely provided ventilation space andinsulation for the first floor (as suggestedat Fréteval; Marshall 2002b, 31). Never-theless the relative ease of access to thebasement suggests practical usage and itmay have been a wine-store, as in contem-porary chamber-blocks and, perhaps, thedonjons at Loches and Rochester (Blair1993, 5; Marshall 2002a, 144, 147).7

The first floor was carried on a systemof twenty-eight radial timbers, the socketsfor which remain, probably supported attheir distal ends on a timber pier as inKing John’s polygonal donjon, from1207-16, at Odiham, Hants. (Hislop 2016,39); cf. Marshal’s tower at Kilkenny,where the pier was apparently of stone(Murtagh 2017, 175), as it may have beenin the circular donjon at Llanhilleth, Mon.,and perhaps at Penrice, Glam. (Renn1961, 142). A much simpler system ofradial joists, without a central pier, wasused at Dundrum and Longtown (McNeill2003, 103-4). Rick Turner has suggestedthat the distal ends of the joists at Pem-broke were sawn obliquely, narrowingthem gradually so that they could meet atthe pier (Rick Turner, pers. correspon-dence; see Figs. 36 and 37); Tom McNeillsuggested a more complex arrangementwith a concentric ring of timbers to whichtheir inner ends were attached (McNeill2003, 104). Either would be a very elabo-rate arrangement for something thatwould never be seen by most users of thedonjon. And while the timbers were0.25m square, and would support a sub-stantial floor surface, the original wall-




Fig. 37 – Section, facing east, through the donjon at Pembroke.


plaster terminates just above the top of thejoists, showing that there was only onelayer of floor-boarding. The flooring arrangement at second-floorlevel was revealed in July 2018, when theinterior of the donjon was scaffolded forremedial work. The sockets are somewhatweathered at this level, but are around0.20m square and are increasingly obliquetowards the east and west, showing thatthey received the ends of eleven more-or-less parallel joists (Fig. 36), as at Château-sur-Epte (Corvisier 1998(3), Fig. 101) andelsewhere. Two slightly lower sockets, tothe east and west, appear to have receiveda cross-beam, not quite at right-angles tothe joists, providing additional support asin the towers at Clogh Oughter andInchiquin, Co. Cork (McNeill’s Group 1A;McNeill 2003, 100 and fig.). It is difficult to envisage how the upper-most storey was floored. The shallow sock-ets are perhaps radial to the tower, but arevery ruinous and difficult to identify indi-vidually; moreover, they seem never tohave formed an uninterrupted series. Threelie a slightly lower level, but do not form acoherent group. Three surviving corbels, atdiffering heights around the second-floorchamber, possibly carried upright postsagainst the wall-face, into which bracesmay have been jointed to support thesethird-floor timbers. The two fireplaces, which have slightlyrounded backs, are unusual. They are verytall, with segmental heads some 3.5 metresabove floor level (Fig. 38). They look likethey have lost their hoods, being similar inform to the formerly-hooded fireplace atVire (Calvados in Normandy), in the squaredonjon built by Henry I (Châtelain 1983, 22and n. 11) and seemingly remodelled by

King John. However, while corbels atshoulder-height (represented by emptysockets at first-floor level) may have sup-ported lintels, there is no toothing for thehood itself, which survives at Vire (Fig.38). As existing, the fireplaces are withoutclose comparisons;8 are they another Mar-shal idiosyncrasy?

The first-floor, entry-level chamber waslit by two slit-lights, and heated by a wall-fireplace. It may have been the waitingarea or anteroom for a second-floor audi-ence chamber, as it possibly did in therectangular donjon at Richmond, Yorks.(Marshall 2016, 165) and, in an invertedsequence, in the great tower at ChâteauGaillard (Marshall 2002b, 32); like mostcylindrical donjons, Pembroke lacks theforebuilding suggested to have housed thewaiting area at e.g. Castle Rising, Doverand Rochester (Dixon 1998, 50; Marshall2002a, 142, 146; Marshall 2002b, 28-9).9Alternatively, the first floor was perhapsitself a reception chamber, at a lower ordifferent societal level from a second re-ception chamber on the floor above, as atRochester and, possibly, Richmond (Mar-shall 2002a, 144, 146; Marshall 2016,165), perhaps presided over by an official(or junior family member) as suggested atHedingham (Dixon 1998, 55). The second-floor chamber is lit by anelaborate twin-light window of high qual-ity, embellished externally with carveddetail (see above) and showing window-seats. There are two additional slit lights,a wall-fireplace and the external doorwaynoted above. It may represent a privateaudience chamber, or perhaps an upperreception room as at Hedingham, for useby the earl should he choose to visit. Ineither role, the chamber would be a fittingenvironment in which tenant lords and



guests could be received, providing a grandsetting for formalities like receiving hom-age, the award of grants and charters, anddiplomatic transactions (see Marshall2002a, 142; Marshall 2016, 164).10 An ap-pearance balcony will have enabled someof these formalities to be witnessed andconfirmed by a wider audience. In the ab-sence of a niche or recess, it is possible thatthe earl’s chair was situated in front of thewindow, a favoured location which placedvisitors at a psychological disadvantage(see Marshall 2002b, 31). Lying beneath the dome, the uppermostchamber had the best lighting with a sec-ond large window, of high quality, and fourslit-lights. It was unheated and, like forexample Hedingham (see Dixon 1998, 55),shows no arrangements for sleeping ac-commodation. The dome was doubtlessintended as the visual centrepiece, and itsunderside embellished with a decorativepainted scheme. However, there is no pisci-na and, while the window is correctly ori-ented on the west side, there is nocorresponding window on the liturgicaleast side to light an altar: the chambertherefore was probably not a chapel ororatory.11 It may then be an alternativecandidate for the role of audience chamber,while the lack of a fireplace militatesagainst prolonged episodes of use. Never-theless, it may have been reserved for theentertainment of a special guest – perhapsthe king, should he visit – as suggestedelsewhere in Britain and France (Marshall2002b, 30). The window gave views to-wards Pembroke Priory, at Monkton, thekind of ‘seigneurial marker’ that guideddonjon window planning elsewhere (Mar-shall 2016, 163; also see Wiles 2014, 183);the arrangement at Pembroke, in which theuppermost chamber window took advantage

of the view, while the window in the more‘public’ chamber below faced the castlebuildings, can be seen in other donjons(Gregory and Liddiard 2016, 149).12

Running around the summit of thischamber is an offset which supported thetimber centring used in the constructionof the dome, the underside of which pre-serves the impressions of plank shutteringin its plaster finish.13 The offset was latertaken advantage of by an inserted floor(see below); two groups of three corbelswhich lie just below it, on opposite sidesof the interior, might have been be associ-ated either with dome construction or thisflooring.The summitThe upper works have undergone consid-erable alteration. Just beneath summit lev-el is a ring of what appear to be socketsfor a timber gallery or hourd. As DavidKing noted, many such sockets are toosmall to take a functional timber and mayoften have been merely for drainage, butat Pembroke ‘the arrangements for drain-ing the wall-walk, which are most effi-cient, are quite distinct from the row ofholes beneath the parapet, each of themcapable of holding a massive timber’(King 1978, 102). Hourding was used systematically in theouter curtain at Château Gaillard (Hislop2016, 160; Mesqui 1993, 327, 334); it isabsent however from the donjon. Unequiv-ocal evidence for hourding is infrequent inAngevin France and Britain, and is bestseen in the donjons at Pembroke andConisbrough (see above), the donjon atRochester, the donjon and towers at Caldi-cot, the inner gate at Corfe (Dorset) and thenorth tower at Stokesay, Shrops. (Davis2016, 257-60; Goodall 2011, 156-7 and



figs.; Hislop 2016, 161-2). The hourding atStokesay, which survives and has beendated to the 1290s (Hislop 2016, 52-3), isdomestic in character and was a perma-nent, rather than temporary fixture, inte-gral with the tower roof structure. Thismay have been normal (see Hislop 2016,51), and is seen elsewhere in, for example,the chamber-tower at Laval (Mayenne),dated 1219-27 (Chollet and Gousset 2012,261; also see Pré 1961, 353-72). We have seen that the donjon at Pem-broke was not intended for defence andperhaps never accessed by soldiers. As inother donjons, this restriction may haveextended to the summit: as height wasequated with status (Marshall 2002a, 143),it follows that the parapet should be themost exclusive – and excluded – zone inthe donjon.14 And recent studies suggestparapets were accessed primarily for theirviews, limited to privileged individualsand guests to whom the lord’s seigneurial‘markers’, like religious houses, could bedisplayed (Gregory and Liddiard 2016,148; McNeill 2006, 123, 126-7).15 It istherefore possible that the gallery at Pem-broke, and elsewhere, was similarly in-tended for non-military uses: those atChâlus-Chabrol (Haut-Vienne) and Salon-la-Tour (Corrèze) were entered throughdecorative doorways from high-statusrooms (Marshall 2002b, 33-4), and some-thing similar may be implied in the circulardonjon at Edwardian Flint, where the tim-ber summit gallery was described, in 1301,as ‘noble and beautiful’ (Renn and Avent2001, 27). Above Pembroke’s sockets is the parapet,which is an accomplished work of somecomplexity, thicker at the merlons than theembrasures. The former are pierced byplunging loops with rectangular basal

oillets,16 perhaps primarily for display asin the Wogan stair turret, where the lowestand most prominent slit-light has an iden-tical, but non-functional oillet. The para-pet nevertheless does not co-exist happilywith the provision for hourding: if anymilitary intent is to be implied, the loopscould not be used once the hourding wasin place while, if the loops were for dis-play only, they would be concealed fromthe viewer behind the decking. Whilelooped parapets may have co-existed withthe hourding-sockets at e.g. Caldicot andChâteau Gaillard (Hislop 2016, 160),Pembroke’s sockets lie well below thelevel of the parapet drainage chutes whichwill have discharged, with damaging ef-fect, over the decking (see Fig. 37). So achange of design may be implied, inwhich the gallery was abandoned in fa-vour of a looped parapet; given that thedonjon may have reached parapet heightby 1207 (see above), this change maycoincide with Marshal’s return from Lein-ster in 1213, after which the new designwas implemented (cf. Murtagh 2017,156).17 The looped upper gallery in thedonjon at Kenilworth (Warwicks.) ap-pears similarly to be an early thirteenth-century alteration or addition and, againas at Pembroke, was probably for displayrather than military use (Renn 2012, 175). An inner wall running inside the wall-walk is built against the flank of the dome,rising up to oversail the parapet as atConisbrough (Brindle and Sadraei 2015, 9fig., 13; Davis 2016, 251); it may belong tothe change in design, although a ‘fossil-ised’ line of slabs, low down in its outerface (see King 1978, 103), is possibly asso-ciated with the original arrangement. Thewall is now truncated, but openings (orembrasures?) may be suggested in illustra-



Fig. 38 – Left: the interior of the donjon at Pembroke, facing northwest, showing first- andsecond-floor fireplaces. Right: the fireplace in the donjon at Vire (Calvados), showingkeystones for the hood.


Fig. 39 – John Speed’s plan of Pembroke, 1610 (National Library ofWales). Inset, top right: the castle as depicted in 1595 (from Owen 1897).


tions from 1595 and 1610 (Fig. 39).18 Theelaborate flagged drainage gully runningwithin this wall (King 1978, 103 and Plate11a; Figs. 36 and 37), which is inaccessiblefrom the outer parapet and not a secondwall-walk, cuts through the surface of thedome and therefore secondary. The spiral stair caphouse butts against theouter parapet, blocking its embrasures, andis clearly also secondary but cannot beclosely dated; some kind of superstructureat the head of the stair may be assumedfrom the start. Other later alteration, possi-bly post-medieval, includes the crude inser-tion of an entry through the dome, atwall-walk level, and a light on the oppositeside. The light preserves the impression ofa timber window-frame, of late form; theentry however shows no provision for adoor (see King 1978, 102; Renn 1968, 39)and cannot therefore belong to a periodwhen the tower was still a functional don-jon. The openings indicate that an addition-al floor had been inserted beneath thedome, taking advantage of the ledge for itscentring and converting it into an atticspace.19 This in turn is probably contempo-rary with the insertion of timber partitionsin all three upper levels (King 1978, 102),which were held in chases crudely hackedthrough the pre-existing plaster, and wouldprovide additional support for the atticflooring. It is possible that the chases maypostdate the removal of the third floor sur-face, as the uppermost runs through thesecond and third floors, though perhaps notuninterrupted (Fig. 37). The central circular ‘turret’, the base ofwhich survives at the summit of the dome,may also belong to this later work. It wasalso depicted in 1595 and 1610 (Fig. 39),but only in vague outline; it cannot beaccessed from the wall-walk and its func-tion remains indeterminate.

It is possible that the flagged drainage gul-ly, mentioned above, also belongs to thislate work. The inner concentric wall mayoriginally have supported a conical roof, assuggested at Conisbrough and ChâteauGaillard and a necessary covering for anytimber dome (see above and Fig. 31). But,while vaulting was frequently roofed in theMiddle Ages, domes were often left ex-posed and there appears to be no provisionfor roofing over the vaulted summits at e.g.Manorbier or Laugharne castles, nor else-where in Pembroke Castle itself. Not all of these later alterations, includ-ing the ground-floor entry, may be con-temporary: the donjon may have beenredundant in its original role fairly soonafter it was completed, as at Chepstow(see Turner, Jones-Jenkins and Priestley2006, 42), with different uses found atdifferent periods thereafter. But a radicalrepurposing of the donjon in the fifteenthor early sixteenth century is suggested bythe broad, curving flight of masonry stepsto the first-floor entry which, thoughlargely rebuilt in the twentieth century,was described in the nineteenth centuryand depicted in antique prints and maps.

APPENDIX 2:The Marshals, Pembroke and Ireland

The records indicate that Chepstow Castlewas a favoured residence for both genera-tions of the Marshal family, for whomPembroke was mainly a source of prestige,as their titular caput, and of revenue.Chepstow by contrast was central to theelder Marshal’s chief power bloc inStriguil, Usk, Gloucester and Dean, andwas the main Marshal port. Like his sons,William Marshal saw little of Pembroke –only two visits are certainly known –whose castle was left to witness the deeds,mainly unrecorded, of its seneschals.



Despite Pembroke’s closer proximity toIreland, moreover, Chepstow appears tohave been the Marshal’s port of choice forthe crossing: while Milford Haven was abusy waterway during the Middle Ages,Pembroke itself was not accessible to sea-going vessels. Its maritime activity wasrestricted to local, coastal traffic throughoutthe historic period, and no quay was presentuntil the nineteenth century.Pembroke and Chepstow

There can be no doubt that William Marshaltreated Chepstow as his principal seat inEngland and Wales, and that it remained thecentre of Marshal power (David Crouch,pers. comm.; Crouch 2006, 49). When not atcourt, in Ireland or active militarily and ad-ministratively, Marshal was recorded eitherin Normandy (until 1204), on his Englishestates, or at Chepstow. The association be-comes most apparent after 1204, but he wasapparently there in March 1194 when heheard of King Richard’s death (Holden et al.2002, 509). He chose to reside there in 1206-7 when he withdrew from King John’s court(Crouch 2015, 11; Crouch 2016, 116), andvisited again in 1217 and 1218 (Pat. Rolls1216-25, 79; Hardy 1833, 314, 370). Mar-shal probably kept his treasury and archivethere (see above), while custody of the castleand lordship may, by 1217, have been vestedin his closest companion John of Earley(Crouch 2016, 171), whose son, anotherJohn, was in control in 1219 (Holden et al.2004, 411). The younger Marshals are fre-quently recorded at Chepstow (Crouch2015, 245-6, 251-2, 299-308, 337-8), whileHenry III visited in 1217 (Hardy 1833, 314;Pat. Rolls 1216-25, 79) and 1232 (Cal. Pat.Rolls 1232-47, 6). This stands in stark con-trast to the limited evidence for any closerelationship with Pembroke. It may also besignificant that in the truce made with Lly-

welyn ap Iorwerth, in 1218, Marshal relin-quished Carmarthen and Cardigan to theWelsh prince in return for Caerleon Castle(Hardy 1833, 356; Holden et al. 2004, 397;Pat. Rolls 1216-25, 143): while the twowest Wales castles were Pembrokeshire’sfirst line of defence against the Welsh heart-lands, Caerleon had posed a threat to thesecurity of Chepstow.20

The elder Marshal crossed to Irelandfrom Chepstow in 1207 (Holden et al.2004, 169-73), and William Marshal IIchose the same route when he becamejusticiar in 1224 (Crouch 2016, 12). It wasthe base for their shipping: ‘galeas et aliasnaves Marescalli . . . ad Striguil’ are men-tioned in the 1230s (Crouch 2006, 48-9;Holden et al. 2006, 36, 152) and, as DavidCrouch points out, ‘it was at Chepstowthat the Marshal fleet which commandedthe Severn estuary was based, the fleetthat would have taken the younger earlMarshal to and from Ireland, as it took hisfather before him’ (Holden et al. 2006,36). Pembroke may then have been a stop-over during Marshal’s crossing of 1200(see above), in order to visit his new lord-ship and commission the donjon, ratherthan his port of embarkation. While Rich-ard Marshal appears to have crossed toIreland after visiting Pembroke in autumn1231 (see Crouch 2014, 396-7), it is likelythat, in 1233 and 1234, he used Chepstowwhere he had spent much of his time.After his death in 1234, Richard’s broth-ers seem to have returned from Ireland viaPembrokeshire (David Crouch, pers.comm.). But while Walter Marshal almostcertainly crossed from Pembrokeshirewhen he visited Leinster in 1244-5(Crouch 2015, 402-03, 415-17, 434-5), hemay have used the port of Haverfordwest,where he is recorded in June-July 1245(Williams ab Ithel 1860, 85).



Pembroke and shippingWhile many of the Marshal tenants in Ire-land also held lands in Pembrokeshire, it isfar from certain that they chose to embarkfrom Pembroke. Along with their officials– and those of King John, 1207-c.121321 –they may, like the Marshals themselves,have used other ports such as Marshal-heldTenby (Pembs.), which had a natural har-bour suitable for sea-going vessels, orHaverfordwest. Pembroke’s borough charter, from 1102-30, mentions an anchorage called the‘Cross’ (‘ad crucem’; Ballard 1913, 169;Cal. Pat. Rolls 1377-81, 106-7; Walker1989, 132, 136-7). Next, the Norman-French epic poem The Song of Dermot andthe Earl, probably composed in the 1190s(Crouch 2016, 123 and n. 40), tells us thatHenry II sailed for Ireland, in 1171, from‘The Cross’ near Pembroke (‘la croiz enmer entra’; Orpen 1892, 188-9). Then in1210 King John assembled his fleet andpersonnel, for his Irish crossing, at ‘Cross-below-Pembroke’ (‘crucem subtus Pen-broc’; Hardy 1844b, 172-7). It appears tohave been the Pembrokeshire historianHenry Owen who first suggested that theCross was not a terrestrial location, butinstead referred to an anchorage in MilfordHaven, downstream from the castle, fromwhich ‘small boats . . . took out passengersand cargo to ships lying in deeper water’(Owen 1897, 317 n. 5). Historically, the most important anchor-age in the Haven, for the town of Pem-broke, was Crow Pool in Pennar Pill, 1.5km west of Pembroke Castle. The Pem-brokeshire topographer George Owenwrote, in 1595, that ‘a barque of 40 or 50tuns may enter this creek at low water andride at anchor at Crow Pool, but no furtherwithout help of the tide’; Crow Pool lay on

the west side of Pennar Pill, near the vil-lage of West Pennar, and within it‘barques of 50 tons may ride safe and ingood anchor-hold at all times in low wa-ter’ (Owen 1897, 545-6). Crow Pool isalso marked as an anchorage on LewisMorris’s chart of Milford Haven, from1748, with a depth of nine to twelve feetat the mouth of Pennar Pill (Morris 1748,15 and plate 24). As the historian RonWalker recognized, the ‘Cross’ and CrowPool are one and the same (Walker 1989,143-4); the latter name may be derivedfrom the Norman-French croiz/croix. All of Pembroke’s maritime traffic ap-pears to have been conducted from thisanchorage, from which, when tides per-mitted, lighters or barges conveyed cargoto and from the town – and, during theMiddle Ages, the castle.22 While HenryII’s charter commanded that ‘all shipswith merchandise that enter the port ofMilford and wish to buy or sell on landshall come to the bridge of Pembroke’, itseems that lighters are meant; larger ves-sels remained at the Cross where they paid‘their lawful customs’ (Ballard 1913, 169;Cal. Pat. Rolls 1377-81, 106-7; Walker1989, 137), making it clear that thesecustoms were collected at sea.23

The identification of the Cross clarifiesthe role of Pembroke, as a port, during themedieval period and later. The town itselfwas inaccessible to sea-going vessels, andnever properly developed as a port: anaccount of 1818 tells us that ‘at low waterthe channel is narrow, intricate, and wind-ing, from Crow Pool to Pembroke, andwithin which is an anchorage for smallvessels; but those of burden are excludedfrom this navigation, and thereby the tradeof Pembroke is checked’ (Pughe 1818,362). Its maritime trade was overwhelm-



ingly coastal and, during the Middle Ages,wine from France was shipped from thetransit port of Bristol, in vessels that liketheir crews were both small and local(Howells 2002, 471). By the thirteenthcentury Pembroke, as a port, had beensuperseded by Tenby; where the annualtolls from Pembrokeshire’s ports were re-corded during the fourteenth to sixteenthcenturies, in most cases none were taken atPembroke ‘because no boats called in thesaid port this year’ (Owen 1918, 112, 127,143, 148). The sixteenth-century ‘WelshPort Books’ record only eight vessels be-longing to Pembroke between 1566 and1603 – all small craft between seven andsixteen tons – when trade with the town ismentioned on a mere six dates (Lewis1927, 71-89, 134, 164, 198, 217, 226, 313);it features far more prominently as a mer-chant seat than a port.24 Pembroke is notlisted among George Owen’s ports andhavens in the county (Owen 1906, 349-50),while his description of Pembroke town, of1595, makes no mention of any maritimeactivity (Owen 1897, 557-8); it is similarlyabsent from John Leland’s account of hisvisit in c.1538 (Smith 1906, 115-17), inwhich Tenby, Fishguard, Carmarthen andKidwelly are by contrast described as ports(Smith 1906, 58, 61, 65). Moreover, no quay or wharfage is shownat Pembroke on John Speed’s plan of 1610(Fig. 39), nor in the wealth of artist’s viewsspanning the late seventeenth centurythrough to 1800 when the area of the pres-ent quay – the only practical mooring sea-ward of the town bridge – was a pool orembayment in which the river lapped upagainst the town walls. A number of views,including the Buck prints of 1740 and1748, depict lighters drawn up here at lowtide (Fig. 40); an inlet beneath the Woganis also shown.

Pembroke had enjoyed something of arecovery after the Restoration, with thegeneral growth of trade in the BristolChannel area (Howell 1987, 28), but Hav-erfordwest increasingly took the bulk ofmaritime traffic during the course of theeighteenth century (Donovan 1805, 336;Howell 1987, 294; Soulsby and Jones1975, 24).25 Pembroke’s trade was still alow-key affair in c.1800, when a quayseems to have been first established, intimber (Thompson 1983, 50 fig.), finallyallowing vessels to unload directly at thetown: in 1805 Edward Donovan hopedthat, as a port, Pembroke ‘might one dayrise to consideration’ (Donovan 1805,336). The present masonry quay was builtin 1818 (Evans 1981, 84), but the narrow,winding channel meant that Pembroke re-mained inaccessible to larger vessels (seeabove) and it was soon eclipsed by Pem-broke Dock (Soulsby and Jones 1975, 24). Walter Marshal begun the lighthouse atHook Point, Co. Wexford, which was stillunder construction in 1247 (Cal. Pat.Rolls 1232-47, 500; Murtagh 2016, 230;Sweetman 1875, 429); its foundation maytherefore coincide with his visit to Lein-ster in 1244-5 (see above). It served theimportant Marshal port of New Ross, andthe River Barrow which was navigable forover 60 miles, connecting the leadingMarshal towns at Kilkenny and Carlowwith the sea. Testifying to the importanceof the waterway, the lighthouse is stilloperational after nearly eight centuries(Murtagh 2016, 223). The Pembroke River is a different matteraltogether. While the inlet through whichit flows is a ria, or drowned valley, theriver itself is a sluggish, meanderingstream which, in sharp contrast to the Riv-er Barrow, extends inland for a mere 1½



miles without any further settlement alongits banks. No lighthouse was thereforethought necessary at, or near Pembrokeduring the post-medieval period, when Mil-ford Haven was otherwise well-suppliedwith navigational aids. All the above fac-tors, taken separately or together, wouldmilitate against the suggestion that Mar-shal, or any its lords, established a light atPembroke Castle.26

Additional referencesPrimary sources

Ballard, A. (ed.), 1913 British BoroughCharters 1042-1216 (Cambridge Universi-ty Press).Calendar of Patent Rolls, Hen. III 1232-1247 (London: HMSO, 1906).Calendar of Patent Rolls, Rich. II Vol. 11377-1381 (London: HMSO, 1895).Cole, G. D. H. and Browning, D. C. (eds),1962 Daniel Defoe: A Tour through theWhole Island of Great Britain, 2 (London:J. M. Dent).Donovan, E., 1805 Descriptive Excursionsthrough South Wales and Monmouthshire

in the Year 1804, and the Four PrecedingSummers, 2 (London: Edward Donovan).Green, F. (ed.), ‘Carmarthen Castle. A col-lection of historical documents relating toCarmarthen Castle from the earliest timesto the close of the reign of Henry VIII’,West Wales Historical Records, 4 (1914),1–71.Lewis, E. A., 1927 (ed.) The Welsh PortBooks, 1550-1603 (London: Cymmrodor-ion Record Series 12).Morris, L., 1748 Plans of Harbours, Bars,Bays and Roads in St George’s Channel(London: Lewis Morris).Orpen, G. H. (ed.), 1892 The Song ofDermot and the Earl: An Old FrenchPoem (Oxford: Clarendon Press).Owen, H. (ed.), 1897 The Description ofPembrokeshire by George Owen of Hen-llys, Lord of Kemes 2 (London: Cymmro-dorion Record Series 1).Owen, H. (ed.), 1906 The Description ofPembrokeshire by George Owen of Hen-llys, Lord of Kemes 3 (London: Cymmro-dorion Record Series 1).

Fig. 40 – Watercolour of Pembroke Castle from the north by Francis Place, 1678. Mudflatsoccupy the area of the present quay (National Library of Wales).



Owen, H. (ed.), 1918 A Calendar of Pem-brokeshire Records 3: Pembroke (London:Cymmrodorion Record Series 7).Pughe, W. O. (ed.), 1818 The CambrianRegister, 3 (London: E. & T. Williams).Smith, L. T. (ed.), 1906 The Itinerary inWales of John Leland in or about the years1536-1539, 3 (London: George Bell andSons).Thompson, M. W. (ed.), 1983 The Journeysof Sir Richard Colt Hoare through Walesand England, 1793-1810 (Stroud: Alan Sut-ton).Walker, R. F. (ed.), 1989 ‘Henry II’s Char-ter to Pembroke’, Bulletin of the Board ofCeltic Studies 36, 132-45.Williams ab Ithel, J. (ed.), 1860 AnnalesCambriae, (London: Rolls Series).

Secondary sources (unpublished)Evans, B. H. (ed.), 1981 ‘Buildings of Spe-cial Architectural or Historic Interest: Pem-broke and Pembroke Dock’ (Cardiff:Welsh Office).Soulsby, I. N., and Jones, D., 1975 TheArchaeological Implications of Redevelop-ment in the Historic Towns of South Pem-brokeshire District (University CollegeCardiff).

Secondary sources (published)Blair, J., 1993 ‘Hall and Chamber: EnglishDomestic Planning 1000-1250’, in G.Meiron-Jones and M. Jones (eds), Manori-al Domestic Building in England andNorthern France (London: Society of An-tiquaries Occasional Papers 15, 1-21).Bouet, P., 2016 ‘Châteaux et résidencesprincières dans le Tapisserie de Bayeux’, inJ. A. Davies, A. Riley, J-M. Levesque andC. Lapiche (eds), 135-46.

Crouch, D., 2006 ‘Chepstow under theMarshals’, in R. Turner and A. Johnson(eds), 43-50.Crouch, D., 2014 ‘Earl Gilbert Marshaland his mortal enemies’, Historical Re-search 87/237, 393-403.

Dixon, P., 1998 ‘Design in castle-build-ing: the control of access to the lord’,Château Gaillard 18, 47-57.Gregory, J. and Liddiard, R., 2016 ‘Visi-ble from afar? The setting of the Anglo-Norman donjon’, in J. A. Davies, A. Ri-ley, J-M. Levesque and C. Lapiche (eds),147-58.Howell, D., 1987 ‘Society, 1660-1793’, inB. Howells (ed.), Pembrokeshire CountyHistory 3: Early Modern Pembrokeshire(Haverfordwest: Pembrokeshire Histori-cal Society), 256-298.Howells, J., 2002 ‘Pembroke’, in R. F.Walker (ed.), Pembrokeshire County His-tory 2: Medieval Pembrokeshire (Haver-fordwest: Pembrokeshire HistoricalSociety), 468-76.Ludlow, N., 2017 ‘Monkton Old Hall: afifteenth-century manorial courthouse?’,Journ. Pembs. Hist. Soc. 26, 9-29.Ludlow, N., in prep. Pembroke: Castle,Town and Priory.Marshall, P., 2016 ‘Some thoughts on theuse of the Anglo-Norman donjon’, in J. A.Davies, A. Riley, J-M. Levesque and C.Lapiche (eds), 159-74.Morgan, O. and Wakeman, T., 1854 Noteson the Architecture and History of Caldi-cot Castle, Monmouthshire (Newport:Caerleon Antiq. Assoc.).Pré, M., 1961 ‘Le château de Laval’, Con-grès Archéologique de France 119, 353-72.

William Marshal, Pembroke Castle and Angevin design - Appendix 2 -References


Renn, D., 2012 ‘Arrow-loops in the GreatTower of Kenilworth Castle: Symbolismvs Active/Passive ‘Defence”, Castle Stud-ies Group Journ. 25, 175-9.Renn, D. and Avent, R., 2001 FlintCastle/Ewloe Castle (Cardiff: Cadw).Rowlands, I. W., 1996 ‘William Marshal,Pembroke Castle and the historian’, Châ-teau Gaillard 17, 151-5.Schellenberg, B. A., 1995 ‘Imagining theNation in Defoe’s A Tour Thro’ the WholeIsland of Great Britain’, English LiteraryHistory 62/2, 295-311.Stacey, N., 2011 Framlingham Castle(London: English Heritage).Turner, R., Jones-Jenkins, C. and Priest-ley, S., 2006 ‘The Norman Great Tower’,in R. Turner and A. Johnson (eds), 23-42.Wood, M., 1983 The English MediaevalHouse (London: Bracken Books).

Notes1 There are however two or three gaps in Mar-

shal’s itinerary in 1213-14, and during winter1215-16, which were potentially long enoughfor a visit. King John spent two weeks atPembroke in June 1210, preparing for hisIrish expedition, but seems to have stayedaboard his flagship anchored in Milford Ha-ven, from which all his acts and letters wereissued (Hardy 1844b, 172-7; see below).

2 As Pamela Marshall notes, ‘we may have tocome to terms with a psyche that was preparedto spend a huge amount of money on a buildingthat was used infrequently, perhaps only once,or even not at all’ (Marshall 2016, 171).

3 The latrines in the curtain at Dundrum, simi-larly thought to have been accessed via abridge from the donjon, were built to serve ahall that was demolished when the donjonwas built (McDonald 2017, 267).

4 The wall-walk was raised by 1.8m at its junc-tion with the Dungeon Tower, when the latterwas added (see Ludlow 1990, 28 fig.).

5 In contrast with the donjon bridge at e.g.Tretower, Brecs.

6 David King speculated that this feature, nowtruncated, might be a postern (King 1978,106), but the external opening is only 0.8mwide, while it is not clear how it might havebeen accessed from the interior (the upperlevels were ‘restored’ in the 1880s). Its over-all form is more consistent with that of alatrine outfall, or a pit-latrine with an externalopening for periodic cleaning.

7 While prisoners were recorded in the donjonsat, for instance, Rouen in early 1203 (Bouet2016, 141-2; Holden et al. 2004, 137), andCompiègne in 1204 (Corvisier 2002, 14),they were presumably of high status like Ran-ulf Flambard, who was held in the WhiteTower, London, in 1100 (Marshall 2016,170) – perhaps not entirely incompatible withthe donjon’s intended use – and therefore notkept in the basement.

8 Fireplaces in pre-existing cylindrical donjons,in Angevin France, were generally either neverpresent or have not survived; the fireplace atLe Brandon is truncated, but corbels suggest itwas hooded (Corvisier 1998(3), Fig. 58).

9 Although the presence of a projecting timberporch at Pembroke is indicated by sockets, itwould not be large enough to act as a waitingarea.

10 The Spanish term for donjon is ‘Tower ofHomage’ (torre del homenaje).

11 Only at this level are the rear-arches slightlypointed, as in the external window surrounds,though a change of masons may be indicatedrather than specific function or status (seeabove).

12 The deer-park at Monkton, however, wasprobably a creation of the fifteenth century(Ludlow 2017, 24).

William Marshal, Pembroke Castle and Angevin design - Appendix 2 - References & Notes


13 Similar impressions can be seen in the win-dow embrasures.

14 Tom McNeill is however of the view that, onthose rare occasions that a castle might comeunder siege, this strict restriction might bewaived (McNeill 2006, 123-4, 126).

15 The Younger Marshal east curtain at Cilger-ran Castle, Pembs., from the 1220s, whichprovides extensive views over a wooded riv-er valley, was parapetted on both sides andpossibly roofed (Hilling 2000, 21), possiblyfunctioning in a similar way.

16 Looped parapets had been introduced by the1190s, e.g. at Château Gaillard and Fram-lingham Castle, Suffolk (Hislop 2016, 160;Stacey 2011, 5, 10-11).

17 An account from 1386 mentions ‘timber brat-tices (britagia de meremio) for the protectionof the walls of [Pembroke] castle, built on thewalls of the said castle’ (Owen 1918, 106). Itis difficult to be certain what is meant; thedonjon was, by 1386, far from being a for-ward defence of the castle, and unequivocalevidence for hourding is absent elsewhere.As it appears in an account of repairs, it mayinstead relate to protection from the elementsduring the repair or rebuilding of walling.

18 The former is the earliest known depiction ofPembroke Castle, forming the background tothe initial letter ‘T’ of George Owen’s De-scription of Milford Haven (Owen 1897, 557).

19 The concept of the attic floor sits uneasilywithin the donjon tradition in which heightwas related to status, with the ‘lordliest areasat the top’ (Gregory and Liddiard 2016, 148);the structural evidence confirms that it issecondary at Pembroke.

20 Marshal’s widow Isabel and son Anselmwere both resident at Chepstow when theydied in 1220 and 1245 respectively and, likeWalter Marshal in 1245, were buried at Tin-tern Abbey (Crouch 2006, 47 n. 14; Crouch2015, 36). The biographical Histoire, com-posed in Marshal’s honour in the mid-1220s,was probably written at Chepstow (Holden et

al. 2006, 35-6). Richard Marshal held a grandtournament there in 1234 (Crouch 2006, 49),while Gilbert Marshal spent extensively onthe castle (Coldstream and Morris 2006, 112).

21 It will be argued in Ludlow (in prep.) thatPembroke was in King John’s hands 1207-c.1213 (also see Rowlands 1996, 155)

22 We see the same practice at Carmarthenwhere, by the 1430s, goods were trans-shipped onto lighters at the anchorage knownas Green Castle, some 3km downstream fromthe town (Green 1914, 48), due to the pro-gressive silting described by Leland in the1530s (Smith 1906, 61); the town quay hadhitherto been accessible to sea-going vessels.

23 William Marshal II’s charter to Haverford-west demanded that ‘ships coming with mer-chandise to Milford do not go elsewhere . ..than to Pembroke and Haverford’ (Ballardand Tait 1923, 243), so that customs could becollected at the Cross.

24 Although maritime legislation was re-organisedin 1559, and customs houses were established atPembroke and Carmarthen (Lewis 1927, ix,xv), customs appear to have been collected, asin the twelfth century, at Crow Pool.

25 Daniel Defoe’s claim that 200 ships belongedto Pembroke town in the 1720s (Cole andBrowning 1962, 57) must be treated with greatscepticism; it has been pointed out that, forDefoe, the ‘discourse of improvement’ over-rode ‘all other rhetorics’ and that, particularlyin Wales, he transformed visible deficits intoapparent surpluses (Schellenberg 1995, 300).

26 Not only would a light at the summit of thePembroke donjon be incompatible with its sta-tus and suggested purpose, but it would un-doubtedly have been furnished with a secondspiral stair – separate from that reserved forprivate use – to serve the keepers, irrespectiveof their social standing cf. the segregation, withseparate stairs, seen in the donjons at Appleby,Carrickfergus and Trim (McNeill 2006, 124-6).

William Marshal, Pembroke Castle and Angevin design - Appendix 2 - Notes

William Marshal, Pembroke Castle and Angevin design Neil ...

Documents

Transcript of William Marshal, Pembroke Castle and Angevin design Neil ...