THE PERIOD OF TRANSITION: CYLINDRICAL TOWER-KEEPS
The development of the castle during the twelfth century was governed,
as has been explained, by the methods of attack which its defenders
had to meet. The strong fortresses of the reign of Henry II., with
their stone curtains and rectangular keeps, opposed to the enemy a
solid front of passive strength which defied attack. Sufficiently
provisioned, a small garrison was capable of holding out against a long
blockade, and tiring the patience of the assailants, whose artillery
could make but little impression upon the masonry of the castle. At
the same time, the stone castle with the rectangular tower-keep does
not represent a final point in the perfection of fortification. The
early Plantagenet castles were, on the contrary, merely a departure
from the ordinary type of castle, composed of earthwork and timber, in
the direction of an organised system of permanent stone castles. They
belonged to a transitional period; for, even while they were being
built, improvements upon their most striking feature, the rectangular
keep, were suggesting themselves. During the last twenty years of the
twelfth century, lessons learned by the Crusaders from the traditional
methods of fortification employed in the Eastern empire exercised
a profound influence upon the military architecture of France and
England; and the application of these lessons during the thirteenth
century entirely altered the defensive scheme of the castle, until
the plan of masterpieces of fortification like Caerphilly and Harlech
presented an entire contrast to the plan of defensive strongholds like
Norham and Scarborough.
The first necessity which had governed the development of
fortification was that of enclosing the defended position so as to
present an adequate barrier to attack. The bailey was surrounded
with its palisade; while the palisaded mount, with its wooden tower,
commanded—that is, overlooked—the operations of the defenders within
the bailey, and provided a second line of defence, if the bailey were
taken. As siege-engines increased in strength, stone-work took the
place of stockading. The bailey was encircled by a stone wall with a
certain number of towers on its circumference. It was sometimes divided
into an outer and inner ward by a cross-wall, or, as at Ludlow (96),
a large outer bailey was added to it, which formed a courtyard for
barracks and stabling, and a protection in time of war for dwellers on
the outskirts of the castle and for their flocks. A wall superseded
the palisade round the mount, or a strong tower was built, either
in connection with the mount or on a new site, which commanded the
whole enclosure. Thus the passive strength of the castle was ensured.
But stone walls and towers, however strong, were in themselves an
insufficient protection, unless the defenders could keep themselves
fully informed of the movements of the enemy. It was necessary that
they should be able to command the field in which the besiegers worked,
and especially the foot of the wall or tower on which the attack was
concentrated. The battering-ram, the scaling-ladder, and the mine must
be kept under constant observation. A first step towards this was the
establishment of projecting towers along the wall at intervals. These
flank the wall—that is, the outer face of the wall between them can be
overlooked and protected by bodies of men posted upon the projection on
either flank. At first, however, the system of flanking was far from
perfect; and therefore the next step was in the direction of improving
it, so that every portion of the outer surface of the _enceinte_ might
be covered by the fire of the defenders. This improvement, which we are
about to trace, was effected gradually, (1) by a change in the form of
the flanking defences themselves; (2) by their multiplication at more
frequent intervals. The first of these changes begins to be noticeable
during the later years of the twelfth century: the second was brought
to pass in the first half of the thirteenth century, and led to further
developments in the arrangement of the lines of defence.
The rectangular form of the keep and of the towers on the curtain was
in two respects a drawback to the defence of the castle. In the first
place, the salient angles of the masonry were liable to destruction
by sap and mine. The parallel jointing of the stonework made the
removal of fragments of stone by the bore or pick at these points a
comparatively easy, if still laborious, task. In the second place, the
angles of a tower or curtain, which were thus points of danger, were
precisely the places which the defenders were least able to command
satisfactorily. Each face of a rectangular tower commands the field
immediately in front of it: the range of shot, from the point of view
of each marksman, is in a direction at right angles to the face of
the tower. Strictly speaking, the foot of the whole curtain and its
towers lies within a “dead angle,” as vertical fire from the rampart is
impossible; but the wooden galleries attached to the rampart obviated
this difficulty. But, if the lines of two adjacent faces of the tower
are produced, it will be seen that the space contained by these is
out of the defenders’ range, and within it miners can work securely,
while the main attack is directed against the faces of the rectangle.
One obvious concrete illustration of this is seen at Rochester. When
King John, in 1215, besieged the castle, he directed against it his
stone-throwing engines. Finding that progress by this means was slow,
he set his miners to work. A breach was made in the outer curtain, and
the miners continued their operations on the tower, and eventually,
after much difficulty, broke their way through it.[192] We can see
to-day that the south-east angle of the tower has been rebuilt, and
that the form of the reconstructed turret is round, and not square.
This, no doubt, marks the place where the breach was made: the repairs
are evidently part of the work taken in hand by Henry III. in 1225.
A further weak point in the defences of the castle was the insufficient
flanking of the curtain. In the eleventh century, as we have seen,
flanking towers were discouraged by feudal over-lords, who rightly
recognised the danger which a strongly fortified castle, in the hands
of rebels, might mean to themselves. As time went on, stone curtains
were provided with towers; but these were not many in number, and,
so long as the rectangular form of tower continued in fashion, long
spaces of straight wall were left between the projections. The risk of
providing too many salient angles was probably recognised by military
engineers. From the flanking towers the adjacent part of the wall could
be covered by the artillery of the defence; but, where a long interval
existed between two towers, the wall mid-way was out of effective
range. To protect these unflanked points in time of siege, a body of
defenders would have to be kept on each spot. A twelfth-century castle,
therefore, to be thoroughly defended, needed a large garrison to cover
its numerous weak points. Any attempt to concentrate the defence upon
one threatened spot might lead to the weakening of the defence at other
points, of which the enemy would not be slow to take advantage.
Added to this was an inherent drawback in the normal plan of the
castle. Its wards and keep provided a system of successive lines of
defence, which caused an enemy immense trouble to pierce, but could
not offer a combined resistance to him. In many castles, like Norham
or Barnard Castle, the inner ward and keep were placed at a distant
angle of the enclosure, and were protected from external attack by
steep outer slopes and a river at their foot. In such cases, the wall
of the outer ward offered the first resistance: the inner ward did
not come into action until the enemy had entered the outer ward, and
the defenders had to retire to the inner enclosure. If the inner wall
was breached or stormed, the keep gave the defence its last shelter.
At Château-Gaillard, as has already been described, the chief feature
of the siege was the capture of ward after ward: the defenders, in
despair, did not even attempt to resort, as a final resource, to
the keep. Château-Gaillard was in its own day a model of scientific
fortification. Its fall was therefore a very striking example of the
disadvantage of successive lines of defence, of which only one could
be effectively used at a time. It is true that here and there, as
at Rochester, the keep was placed so near the curtain of an outer
ward that the exterior of the castle could be commanded from its
battlements, and its artillery could be brought into play over the
heads of the defenders of the curtain. At Richmond, the great tower
commands the one side of the castle from which attack was possible,
and was thus placed in the very fore-front of the defence. But such
arrangements were happy ideas which occurred to individual engineers,
and do not imply any systematic advance in the science of defence.
[Illustration: Château-Gaillard; Plan]
The experiences of the earliest Crusaders brought the warriors of
the west face to face with methods of defence far superior to those
employed in England and France. The city-wall of Antioch gave them an
example of a perfect system of flanking defences; and, in the triple
_enceinte_ of Constantinople they saw how successive lines of defence
could be used in co-operation. At Antioch the wall was flanked at
frequent intervals by fifty towers. Each of these, rising above the
curtain, commanded not only its space of intermediate wall, but the
rampart-walk as well. The rampart-walk, moreover, passed through the
towers, which were protected by strong doors. To gain the whole line
of wall, therefore, it was necessary to occupy the towers, each of
which could be converted into a separate stronghold, isolating the
intermediate rampart-walk. The siege was badly conducted, the Crusaders
limiting themselves to a strong position between the city and the
Orontes, and allowing the defenders to hold their communications on
two sides of the city open for some five months. Posts of observation
were eventually established on the two neglected sides; but the actual
capture of the city was due to the treachery of one of the commanders
of the Turkish garrison, who admitted a body of Franks into one of the
towers in his charge. They made their way into seven more towers, and
so gained access to the city.[193] The three walls of Constantinople
surrounded the whole city: each was higher than the one outside it,
so that all three could be used simultaneously by the defenders.[194]
Against such a system of concentric defence, the besiegers were
manifestly at a disadvantage.
These lessons from the east, stimulating though they were, did not
produce their full practical effect for some generations in the
west. Our engineers had to pass through a long epoch of gradual
experiment before they could arrive at a finished system of flanking
or of concentric lines of defence. The traditional mount-and-bailey
plan provided the foundation of the plan of the stone castle. The
traditional importance of the keep as the ultimate place of refuge
dictated the arrangement of ward behind ward, culminating in the great
tower. Meanwhile, the improvement of flanking defences led more and
more to the concentration of engineering skill upon the curtain, so
that the keep gradually took a place of secondary importance. As an
obvious result of further improvement, the keep was dispensed with,
and the whole attention of the engineer was directed to combining the
defences of the castle into a double or triple line of simultaneous
resistance to attack. These steps took time: the transition from one to
the other was effected by no sudden revolution, but by work along old
lines, a work of revision and improvement, until the finished product
formed an almost complete antithesis to the source from which it was
derived.
The earliest signs of transition in England are seen in the
strengthening of the masonry by the reduction and elimination of
salient angles. It is obvious that, if a rounded or polygonal form is
given to a projecting tower, or if the angle of a rectangular tower is
rounded off, a wider field will be commanded by the artillery of the
defence. The new range will be a large segment of a circle radiating
from the centre of the tower, instead of a rectangle in front of each
face. The sectors at the angles within which an attacking party can
work securely will be thus eliminated, and the chances of the success
of a mine will be less. The masonry also will offer much greater
resistance to the battering or boring engines of the enemy. The joints
are no longer parallel, but radiating, so that it becomes much harder
work to force out stones and effect a breach. The obtuse angles of
polygonal towers, with the joints of the masonry in the alternate faces
running in oblique directions to each other, have a much greater power
of resistance than the right angles of the ordinary twelfth-century
tower.
The general use of circular and polygonal forms is first found in
connection with the principal tower of the castle, the keep. The
main object was at first, no doubt, the greater cohesion imparted
to the masonry: the scientific advantages, from the point of view
of artillery, probably were not realised till later. In France the
cylindrical donjon appeared at an earlier date than in England: that
at Château-sur-Epte (Eure) is said to have been begun in 1097.[195]
The tower of Houdan (Seine-et-Oise) is a cylinder flanked by four
cylindrical turrets: it was built during the reign of Louis VI.
(1108-37),[196] and the form shows that the builders looked, not merely
to the strength of the masonry, but to the reduction of the enemy’s
chances of successful attack. The majority, however, of such donjons
in France belong to the second half of the twelfth century and the
beginning of the thirteenth, and were contemporary with our rectangular
towers. But the engineers of Henry II., to whom we owe so many of our
stone keeps, were certainly acquainted with the possible benefits of
forms other than square. The keep of Orford in Suffolk was probably
built between 1166 and 1172,[197] and is therefore earlier in date than
many rectangular keeps.
[Illustration: Conisbrough; Keep]
Internally, it is cylindrical; externally, a polygon of twenty-one
sides, with three very large rectangular turrets projecting from it.
It has a basement and two main floors, and is entered by a two-storied
fore-building, which forms a southward continuation of the eastern
turret. The sloping base of the tower is continued round the turrets,
and greatly strengthens their angles; while the turrets themselves
are so placed as to flank the whole tower and fore-building very
effectively, and to provide additional room in the interior. This
combination of the rectangular and polygonal forms is, for its date, an
unique departure from the ordinary type of English tower-keep. But it
must be remembered that the shell-keep on the mount usually took the
form of a cylindrical or polygonal wall strengthened by buttresses; and
at Orford, where the tower appears to stand upon the base of a levelled
mount, we may have a conscious adaptation of this form to the heavier
and loftier tower. At Gisors (Eure) the older donjon was an octagonal
tower, built on a mount, and surrounded by a circular wall. The tower
was probably built by Henry II. between 1161 and 1184,[198] within
the somewhat earlier shell, and took the form which was best suited
to the artificial soil on which it stood. But there are at least two
instances of English rectangular keeps in which a slight departure from
the normal form was made for obvious purposes of additional strength,
without reference to an artificial site. At Newcastle the north-west
turret is octagonal, with very obtuse angles. In the small tower of
Mitford, on the Wansbeck above Morpeth, the north wall is built with an
obtuse salient angle, so that the tower forms an irregular pentagon.
The date of this tower cannot be fixed with certainty, but it probably
belongs to the second half, at any rate, of the twelfth century; and
it can hardly be doubted that the object of this peculiar device was to
give the defenders better command of the angles of the tower which were
exposed to attack from the inner ward.
[Illustration: Conisbrough; Keep. Plans]
Somewhat later than these is the noble cylindrical keep of Conisbrough
(166), which is attributed to Hamelin Plantagenet, a natural brother of
Henry II., and husband of Isabel, heiress of William, earl of Surrey.
Hamelin died in 1201: the tower was built, as the architectural details
show, during the last quarter of the twelfth century. It is a regular
cylinder; and to its circumference are applied six bold buttresses,
which narrow slightly outwards, and rise above the parapet in turrets.
The whole is built of dressed stone in large rectangular blocks, the
fine condition of which, after more than seven hundred years, is
extraordinary. The construction is unusually solid: the thickness of
the wall in the basement exceeds 20 feet. On the first floor it is
just under 15 feet: in the two upper floors it is reduced by internal
off-sets, until, at the rampart level, 75 to 80 feet above the ground,
it is 12½ feet. In addition to this the buttresses, which project
9 feet at the basement level and 8 feet above, are not used, like the
turrets at Orford, to contain additional rooms, but are built solid.
The chapel, however, was formed by constructing a chamber in the
eastern buttress upon the third floor.
[Illustration: Conisbrough; Fireplace]
The tower of Conisbrough, like that of Orford, was intended for
residential as well as defensive purposes; but light and comfort were
sacrificed to military necessities. The entrance, as usual, was upon
the first floor, but there is no trace of any fore-building, nor is
the original means of approach at all clear. The basement was simply
a domed well-chamber and store-room: the only approach to it from the
first floor was an opening in the centre of the guard-chamber, over
which was probably the windlass by which buckets were lowered into the
well.[199] The first floor was a guard-chamber: there were no windows,
and the only means of admitting daylight was through the open door on
the far side of the passage through the wall. On the right-hand side of
this passage, a curved stair mounts through the thickness of the wall
to the second floor, which it enters by a landing in the embrasure[200]
of a loop on the north side. This floor was the hall of the keep.
There is a large fireplace (168) in the west wall, with a spreading
chimney-breast, and a lintel of joggled stones resting on triple
shafts with carved capitals. In the wall between the fireplace and the
entrance is a rectangular recess, containing a small sink, which was
drained through the wall. There are two windows, the loop close to the
entrance, and a double window opening to the south-east. The embrasures
are barrel-vaulted: that of the double window has a stone bench on all
three sides, and stands three steps above the floor of the hall. This
window was not glazed: the upright between the two rectangular openings
has at the back a rounded projection, through a hole in which the bolt
of the shutters passed, and the fastening was further secured by a
wooden draw-bar. On the north-east side of the hall a winding passage
with two turns and a flight of steps leads through the thickness of the
wall to a garde-robe.
To reach the third floor, the hall had to be crossed to a recess in
the direction of the south-west buttress. From this point a curved
staircase mounted through the wall to the embrasure of a loop in
the south-east face of the third floor. The apartment on this floor
contained a smaller fireplace, immediately above that on the second
floor, and treated with similar architectural ornament. The flue of
the lower chimney runs up through the wall behind that of the other:
the common chimney-top projects from the rampart-walk above. There is
also upon this floor a trefoil-headed recess with a sink. There are two
windows, the loop in the south-east face, and a double opening, similar
to that below, looking south. This room corresponded to the “great
chamber,” which is found in the larger houses of the middle ages. On
its east side the chapel, an irregular hexagon, vaulted in two ribbed
bays with a transverse arch between, was constructed in the eastern
wall and buttress. The details of its beautiful capitals, like those of
the fireplaces, show elementary foliage of the water-leaf type, such as
is found in the chapel of the tower at Newcastle (152). Chevron is used
in the stilted transverse arch and round the outside of the arch of the
loop at the east end. The quatrefoil openings north and south of the
chancel bay, and the trefoil-headed _piscinae_ in the same walls, are
of an advanced transitional character; and, by comparing these details,
a date approximating to 1185-90 may with some certainty be given to the
tower. In the north wall of the chapel a doorway leads into a small
vestry or priest’s chamber, lighted by a loop. The stairway to the
rampart-walk mounts through the wall above this chamber, and its head
is above the western bay of the chapel. It is entered from a recess in
the north-east wall of the second floor, and from this recess there is
also a zigzag passage to a garde-robe, the seat of which is corbelled
out in the angle between the north-east buttress and the north wall of
the tower. The two lower stairways and the two garde-robe chambers are
each lighted by a small loop.
In the roomier arrangement of the keep at Orford, the stair is a vice
in the turret or buttress to which the fore-building is annexed. The
chapel is upon the first floor of the fore-building, and, being on a
level of its own, not corresponding to the levels of the tower, is
approached from the stair by a separate passage. The entrance to the
chapel is a doorway on the left of this passage, which is continued
through the south-east wall of the tower to a priest’s room in the
south turret.
The defensive side of the arrangements at Conisbrough must now be
considered. The tower stands close to the north-east corner of a
large bailey, the shape of which follows that of the knoll on which
it is built: the north segment of the tower, with the two adjacent
buttresses, continues the line of the curtain; but five-sixths of the
circumference, with four of the buttresses, are within the enclosure.
On the north and east sides the steepness of the hill made access
nearly impracticable, and the natural point of attack was from the
south and south-west. The position of the keep is at the point furthest
removed from attack, and the capture of the inner ward, as will be
seen in a later chapter, was rendered very difficult by a well-guarded
approach.[201] The tower stood on higher ground than the rest of the
ward, and the entrance, on the south-east side, was sheltered by the
east curtain. The south and south-west faces were fully exposed to an
attack from the inner ward, and it was on this side, therefore, that
the defenders needed full command of the sides and base of the tower.
Accordingly, when we mount to the rampart-walk, and examine the tops of
the buttresses, we find that the two which are upon the north curtain,
and were not exposed to attack, contain cisterns. The two on either
side of the main entrance were not necessary for flanking purposes, as
the entrance itself would be defended by some kind of platform in time
of siege. One, therefore, above the chapel, was employed as a house for
carrier-pigeons; while the other contained an oven, in which stones
and arrows could be heated. The remaining two buttresses are raised
platforms which effectively flanked that part of the circumference
which was otherwise insufficiently guarded, and lay open to catapults
and mining operations. The spreading base of the tower and buttresses
served further to keep the battering-ram and bore from direct contact
with the main wall of the tower, and improved the flanking position of
the defenders; while missiles dropped from the summit upon this talus
or sloping surface would rebound upon the enemy with deadly effect.
Above the talus the solidity of the main wall defied the force of
catapults. These engines, however, had increased in strength and range,
and it was no longer safe to give light to the tower in the somewhat
lavish method adopted by the engineers of some of our large keeps. At
Conisbrough, as we have seen, the walls of the first floor, save for
the entrance passage, are absolutely solid. The loops in the upper
floors are very few in number, and the one on the most exposed face
is almost concealed by a buttress. The double window on the second
floor is immediately over the main entrance, on a side which it would
be difficult to command with a large siege-engine. That on the third
floor is placed upon an exposed face, but would probably be out of
range.[202] The garde-robe vents are on the side where the tower
crosses the line of the curtain.
In time of siege the larger windows would be shuttered and barred. The
defence would be conducted from the top of the tower, while a body
of the garrison would be told off to protect the main entrance. The
whole summit would be utilised. The defence was not confined, as in a
rectangular keep, to the rampart-walk; but there was a rear-wall to
the walk, through which openings probably gave access to a covered
round-house above the third floor. To this room, which, to judge
from contemporary instances, had a conical roof, arms and missiles
could be hauled up, through trap-doors in the floors below, from the
guard-room and the store-chamber in the basement. There was no vaulted
roof in the tower above the basement, so that the flat roof could
not be used as a platform for catapults. There is no indication that
hoarding was employed outside the rampart. The tower and its buttresses
were finished off with a battlemented parapet in the usual way; the
buttresses, as has been shown, were so constructed and so near together
that additional wooden defences were practically unnecessary.
[Illustration: Etampes; Donjon. Plan]
[Illustration: GOODRICH CASTLE: round tower with spur at base]
[Illustration: GOODRICH CASTLE: buttery hatches]
In France the treatment of the donjon was pursued with more variation
than in England. To the middle of the twelfth century belongs, for
instance, the donjon of Etampes (Seine-et-Oise), which takes the form
of a quatrefoil (172). The donjon of Provins (Seine-et-Marne) is of
much the same date, the ground-plan forming an octagon flanked by
four cylindrical turrets. Although both these towers have analogies
in England, they were constructed nevertheless by French engineers
at a period before even the rectangular tower had become common with
ours.[203] They are also only two out of many diverse experiments. The
cylindrical form, however, commended itself to the builders of the
finest French examples. Château-Gaillard (163) follows closely upon
Conisbrough in point of date, having been begun by Richard I. in 1196.
The donjon is not, as at Conisbrough, a tower to which the line of a
somewhat earlier curtain has been adapted, but is part of a homogeneous
scheme of fortification. The site of the castle is the top of a very
steep and almost isolated hill on the right bank of the Seine: the west
slope is a precipice, and the only practicable attack could be made
from the ridge joining the hill to the high ground on the south. The
donjon is set so that its west face projects from the curtain of the
inner ward, upon the very edge of the precipice. The interior forms a
regular cylinder, and the west face is a segment of a circle. On this
side the solidity of the masonry is increased by a tremendous outward
slope or batter, the whole height of the basement and adjacent
curtain. Towards the inner ward, however, the cylinder is strengthened
by a covering spur, also battered, so that, while the interior of
the castle was commanded from the rampart, the tower offered to the
besiegers an angle of immense thickness and strength, immediately
opposite the gateway of the inner ward. A possible prototype in
France of this form of defence is the donjon of La-Roche-Guyon
(Seine-et-Oise), higher up the Seine, where the spur covers about a
quarter of the circumference of the tower. Philip Augustus adopted
the same device a few years later in the White tower at Issoudun
(Indre).[204] It is seen at Goodrich (174), Chepstow, and elsewhere.
[Illustration: Château-Gaillard]
As the upper portion of the tower of Château-Gaillard is gone, its
internal arrangements are difficult to decipher. It was purely a tower
of defence; but the inaccessible nature of the west side allowed of
large windows being made in that face upon the first floor. There was
probably a low second floor, above which was the roof and rampart-walk.
The rampart was defended with the aid of a device, unusual at the
time, although very general at a later period. The sides of the tower
within the ward were furnished with narrow buttress projections above
the battering base, which gradually increased in breadth as they went
higher. These divided the face of the tower into a series of recesses
spanned by low arches, on the outer face of which the parapet was
carried. The top of each recess, between the parapet and the wall,
was left open, so that the defenders could use the holes for raining
down missiles upon their opponents. Such holes, formed by corbelling
out a parapet in advance of a wall or tower, are called machicolations
(_mâchicoulis_),[205] and gradually superseded the external gallery
of timber. Holes in stone roofs for the same purpose are found at an
earlier date, as in the fore-building at Scarborough; and, as early as
1160, they appear in connection with the parapet of a donjon at Niort
(Deux-Sèvres).[206] The general tradition is that they were invented
by the Crusaders in Syria, where wood for hoarding was not easily
obtained; and this is probably true.[207] They appear in a state of
perfection, which testifies to a long course of previous experiment,
at the great Syrian castle of Le Krak des Chevaliers (176), begun in
1202. But hoarding continued in use in Europe long after the building
of Château-Gaillard, and even the donjon of Coucy (Aisne), by far the
finest of all cylindrical donjons, was garnished with timber hoarding
carried on stone corbels—an interesting example of the transition from
one form of defence to another.
[Illustration: Le Krak des Chevaliers]
[Illustration: Coucy]
The cylindrical form of donjon was brought to perfection in France
under Philip Augustus (1180-1223). At Gisors, which came into
his hands in 1193, he built a new circular tower on the line of
the curtain, which superseded Henry II.’s octagonal tower on the
mount. His fortification of Gisors led directly to the building of
Château-Gaillard by Richard I., to cover the approach from French
territory to Rouen.[208] But in 1204 the capture of this great
stronghold delivered Rouen into Philip’s hands; and in 1207 he built
the donjon, now known as the Tour Jeanne d’Arc, at Rouen. Here we meet
with the tower vaulted from basement to roof, with a strongly defended
entrance at the level of the ward in which it stands, of which the most
perfect example is found at Coucy.[209] Coucy, the work of a powerful
vassal of the crown of France, represents a degree of scientific
fortification to which none of our cylindrical donjons attains. The
castle was constructed, like Conway at a later period, in connection
with the defences of a walled town.[210] It consists of two wards, a
large outer ward or base-court and an inner ward of irregular shape,
with four straight sides of unequal length and round towers at the
angles. In the middle of the east side, between the two wards, is the
donjon (177), a cylinder of some 200 feet high—90 feet higher than the
tower of Rochester. It stands isolated from the curtain of the inner
ward, from the line of which about one-third of its circumference
projects, and is surrounded by a ditch, originally paved with stone.
To this ditch there was no external access. On the outer edge of the
ditch, joining the east curtain of the inner ward at two points, was
a strong wall or _chemise_. Outside this was the ditch dividing the
inner ward from the base-court. Within the inner ward, a low wall took
the place of the _chemise_ of the donjon, and access to the tower was
provided by a bridge across the stone-flagged ditch. The bridge was
worked by a windlass, and, when not in use, remained drawn up on the
threshold of the tower.
The donjon of Coucy is built in three stages, and has a large
apartment, originally vaulted, on each floor. There is no basement
chamber below the level of the entrance. In order to facilitate
vaulting the various floors, each chamber was planned with twelve
sides, lofty niches being left between the abutments of the vault.[211]
Without giving a detailed description, we may notice the points in
which this great structure resembles and improves upon the tower of
Conisbrough. (1) The isolation of the tower, defended by its own ditch
and, towards the field, by its own curtain, makes an entrance on the
ground floor possible. In this respect, the builders of Coucy followed
the example of Philip Augustus at the Louvre and at Rouen. (2) The
defences of the entrance are more elaborate than at Conisbrough, where
the doorway was closed merely by a strong wooden door, reinforced
by two draw-bars, and a straight passage led into the guard-room on
the first floor. At Coucy there was a similar door, but in front of
it was an iron portcullis, worked from the first floor of the tower,
and sliding through grooves at the back of the jambs of the doorway.
The portcullis was defended further by a machicolation or open groove
in the floor above. The entrance passage behind the wooden door was
closed by a hinged grille at the entrance to the guard-room. (3) The
stair, as at Conisbrough, was on the right of the entrance passage,
but, instead of following the curve of the wall, was a vice, which
led straight to the roof, communicating with the two upper floors on
the way. The device adopted at Conisbrough, by which the stair ends
at each floor, and, in order to ascend further, the floor has to be
crossed, was adopted in the lesser towers at Coucy,[212] but not in the
donjon. The Conisbrough method has the advantage, very desirable in a
tower, of keeping the approach to the roof under direct observation
throughout its entire distance: we find it used in the stairs of the
rectangular keep at Richmond. (4) The tower of Coucy, as already
noticed, was defended by a lofty parapet, pierced with arches, which,
in time of siege, gave access to an outer wooden gallery supported
by stone corbels.[213] The form of the corbels is that which became
general in later times: each is composed of four courses of stone
projecting one above the other, with their outer ends rounded. (5)
The well at Coucy was in one of the niches between the abutments of
the ground-floor vault. (6) There are garde-robes at Coucy on the
left of the entrance-passage, and in a similar position at the entry
to the first floor. (7) We have seen that at Conisbrough arms were
probably transported from the basement to the roof through a series
of trap-doors in the floors. At Coucy there was a circular opening
left for this purpose in the crown of the vault of each floor. (8)
The solidity of the tower of Coucy is emphasised by the absence of
large windows, even more noticeable than at Conisbrough; and, although
the tower contains fireplaces, its purely defensive character is
unmistakable. It provided accommodation for an enormous garrison, but
for residential purposes, it would have been uncomfortable to the last
degree. It contains no trace of a permanent chapel: when the tower was
in use, an altar might have been set up in one of the niches on the
first floor; but the regular chapel was in the inner ward, and was
connected with the domestic buildings.
In the walls of the tower of Coucy can still be seen the holes which
served to attach the scaffolding during construction. The spiral course
which they take shows that the scaffolding, rising with the tower,
formed an inclined plane of a moderate slope, up which the necessary
materials could be wheeled. The advantage of a cylindrical tower from
this point of view is obvious. Another structural feature is the
provision of gutters for the drainage of the roof in the stonework at
the back of the vault-ribs of the second floor. The absence of any
effective provision for draining the centre of the roof at Conisbrough
points to the probability that it was sheltered, as already explained,
by a conical roof of its own.
[Illustration: Pembroke]
The introduction of the cylindrical donjon in England coincides with a
period at which the keep was already beginning to disappear from the
castle. The principal examples, which may be attributed to the early
years of the thirteenth century, are on the frontier and in the south
of Wales. Chief among them is the fine tower of Pembroke (180), which
was probably built by William Marshal, earl of Pembroke and Striguil,
about 1200. The castle of Pembroke was of great importance, owing to
its situation upon an arm of Milford haven,[214] and its command of the
passage to Ireland. The keep was probably the first completed portion
of the present castle, the stone-work of which, as it stands to-day,
is very largely of the late twelfth and early thirteenth century.[215]
It is a round tower, with a basement and three upper floors, standing
just within, but not touching, the curtain which divided the inner and
higher from the outer ward. The height is 75 feet; the floors were
of wood, but the uppermost stage was vaulted by a dome, which still
remains, rising in the centre of the tower above the rampart-walk. The
stair is a vice in the west wall, from the basement to the summit: the
main entrance was upon the first floor, but there is also a basement
entrance, which seems to have been pierced not long after the building
of the tower. The whole structure batters upwards, and the walls are
slightly gathered in at each stage on the outside, a method the reverse
of that pursued at Conisbrough: the masonry is roughly coursed rubble.
On each of the first and second floors there is, towards the inner
ward, a two-light window with pointed openings, the spandrils between
which and the enclosing arch are pierced with plate tracery. The third
floor was lighted by windows pierced in the dome.[216] Commanding, as
it does, the whole interior of the castle, this tower is remarkably
grand in situation; and its thick walls offered considerable resistance
to artillery. It shows, however, no advance upon the defence of
Conisbrough. The rampart-walk is narrow, and the dome in the centre
prevented the employment of the roof as a platform.
[Illustration: Pembroke; Plan]
The cylindrical donjon in England and Wales was simply an experiment
attempted here and there, as an improvement upon the rectangular tower,
but was never carried to the general perfection which it attained in
France. Its isolation at Coucy, upon the outer face of the inner ward,
protected by its own inner ditch, and covered by a strong curtain of
its own, are signs of a perfection of engineering skill to which our
builders did not attain. In one case, at Flint, we find a round tower
which is isolated within its own ditch at one corner of the castle,
but stands outside the main wall, and had no separate curtain of its
own.[217] The plan strongly suggests a mount-and-bailey fortress, the
isolated tower occupying the site of the mount, and the bailey walled
in, leaving the moat, which was marshy and was filled with water at
high tide, clear. The construction of this keep is peculiar: it is
composed of an outer and inner circle of masonry, with barrel-vaulted
passages between the two. Its actual date is unknown.[218] But, as
a rule, where the keep stands upon the outer line of defence, it is
joined by the curtain of the bailey. Thus at Caldicot, near Chepstow,
the castle is simply a mount-and-bailey enclosure surrounded by a stone
curtain of the thirteenth century. The keep is a round tower at one
corner, standing upon the partially levelled mount; and the curtain
crosses the ditch to join it on both sides. At Conisbrough, where the
keep was on the line of the curtain; at Pembroke, where it stood just
within the line, there was no ditch round it: the high ground on which
it was placed seems to have been thought a sufficient protection.
There are, however, a few round towers which, although they have not
their own curtain in the sense of Coucy, are yet within defences of
a peculiar nature, and therefore stand in a class apart. The most
remarkable of these is Launceston, where the tower stands upon the
summit of a lofty artificial mount of early Norman origin, and is
approached by a steep and well-defended stair, ascending the face of
the mount to the main entrance. Round the outer edge of the mount
remain the lower courses of a stone wall, concentric with the keep.
Within this is another and higher circular wall, which was crowned by
a rampart-walk, approached by a stair in the thickness of the wall, to
the left of the entrance. Inside this enclosure is the tower itself,
which now consists of a basement and a ruined upper floor. The narrow
space between the tower and the encircling wall was evidently roofed
over at the height of the first floor of the tower: holes for joists
still remain.[219] This double circle of masonry recalls Flint, where,
however, the intermediate passage was vaulted, and the outer circle was
probably the whole height of the tower.[220] Flint does not possess
the low outer wall which existed at Launceston. The nearest analogy to
Launceston is at Provins (Seine-et-Marne), where the octagonal keep
has its own outer curtain, and is composed of an outer octagon with
cylindrical turrets at the angles, commanded by an inner octagon rising
two stages higher. The upper stage at Provins is surrounded by a lofty
crenellated wall, on which rests a conical roof.
[Illustration: Dolbadarn]
Another case is the keep of Tretower in Breconshire, which stands on
a slightly elevated site near the confluence of the Rhiangol with the
Usk. Here the arrangement is very curious. The keep, a round tower
with a basement and three upper stages, stands within the ruins of
an approximately rectangular enclosure. This enclosure bears a close
resemblance to the outer wall of a rectangular keep, but has two
octagonal projections from the south face, one of which contains a
vice, and the other a large fireplace. The tower itself seems to be
somewhat earlier than the year 1200: the fireplaces on the first and
second floors have architectural decoration recalling that of the
fireplaces at Conisbrough, shafts with capitals carved with foliage of
a very elementary kind. The solution which suggests itself is that a
rectangular tower, of a somewhat original plan, was begun and raised
to a certain height, and that the builders then changed their minds,
built a circular tower within the unfinished keep, and left the outer
walls to serve as a curtain for the new structure.
The keep at Tretower, in its ordinary features, may be compared with
the tower of Bronllys, only a few miles distant, on the other side of
the pass through the Black mountains, at the southern foot of which
Tretower stands. This tower also seems to be a work of the end of the
twelfth century, but its architectural details are much plainer: both
seem originally to have been between 70 and 80 feet high, and each
contained a basement and three floors. Each has a battering base, and
above this the wall at Tretower batters slightly to the summit; the
diameter of Tretower exceeds that of Bronllys throughout. The original
entrance in each case was on the first floor, from which at Tretower
a vice led to the top of the building. The basement at Tretower had
its separate stair in the wall opposite the entrance. At Bronllys the
basement has a pointed barrel vault, and was entered by a stone stair
and ladder from a trap-door in one of the window recesses of the first
floor. The stair from the first floor to the second opened from another
window recess, and curved through the wall, as at Conisbrough; there
was, as also at Conisbrough, a separate stair to the third floor. The
wall of the basement at Bronllys has been broken through in two places,
and in one of these a hollow in the wall has been disclosed, in which
originally a great beam was inserted to give coherence to the masonry.
The same feature is seen in the outer building at Tretower. This device
was frequently employed in the construction of medieval walls, but its
traces are not often so clearly seen.
[Illustration: Dolbadarn; Interior]
One feature of the tower of Bronllys is that, like that of Caldicot, it
stands upon an artificial mount, which occupies the ordinary position
of such earthworks, at the head of the enclosure. The more roomy,
but lower, tower at Hawarden, the upper floor of which is internally
an octagon, almost surrounded by a mural passage, is built upon a
lofty mount. At Skenfrith in Monmouthshire the tower, nearly equal
to Bronllys in diameter, but not higher than Hawarden, stands upon a
very low mount, and is placed in an isolated position, nearly in the
centre of a trapezoidal enclosure. Here the lowness of the mount and
the absence of indications of a normal earthwork plan suggest that it
was raised to strengthen the foundations of the tower, and is not the
mount of an earlier castle. The knoll, on the other hand, on which
the round tower of Dolbadarn (183) stands, between the two lakes at
the foot of the pass of Llanberis, is natural. The details of this
tower are very plain, but it was probably built during the thirteenth
century. There is no trace of any castle in connection with this small
military outpost, which, like the not far-distant rectangular keep of
Dolwyddelan, on the eastern slopes of Moel Siabod, bears some analogy
to the “pele-towers” of the north of England, and may have been built
by a Welsh chieftain upon an English model during the reign of Henry
III.
[Illustration: York; Clifford’s Tower]
None of the towers in England and Wales mentioned in this chapter have
the inner spur which has been noticed as characteristic of French
towers. It appears, as has been said, at Goodrich and Chepstow. Other
instances are a tower in the outer curtain at Denbigh, and the spur on
the inward face of the great tower at Barnard Castle. Here the work
is not earlier than the time of Edward II., and the tower is little
more than a large mural tower added to a large shell-keep standing on
a high rocky point. The spur here is a half pyramid, the apex of which
dies away in the face of the tower. Of an octagonal tower we have one
example at Odiham in Hampshire, which may be of the end of the twelfth
century. This has the feature, anomalous for so early a date, of angle
buttresses which project 4 feet, but are only 2 feet broad.
[Illustration: Berkeley Castle; Plan]
Of donjons which were built in England during the reign of Henry III.,
the most interesting, by virtue of their plan, are those of York
and Pontefract. The tower of York (185), raised upon the mount of
the northern of the two castles, was built possibly about 1230, and
assumed the quatrefoil shape which is found in France at Etampes. This
keep, presumably because it is built on a mount, is usually called a
“shell”; it was, however, a regular tower, and the entrance, in the
angle between two of the leaves of the quatrefoil, is guarded by a
rectangular fore-building, on the first floor of which was the chapel.
As at Etampes, the quatrefoil plan is preserved internally, but the
angles formed by the meeting of the four segments are chamfered off:
there was no vaulting, as at Etampes, but the floors were of wood.
A quatrefoil plan was also adopted at Pontefract, with some slight
variation, owing to the irregular shape of the rocky mount. This keep
is in a state of complete ruin, although some idea of its former shape
may be gathered from a bird’s-eye view preserved among the records of
the duchy of Lancaster.[221] We can see, from what is left, that it
was not built upon the top of the mount; but that, on three sides, the
mount was enclosed by walls of revetment,[222] which formed the base
of the segments composing the quatrefoil. This process recalls the
walling-in of the mount at Berkeley, where, however, the lower part of
the mount was left, and the space between the slope and the wall filled
in with earth. At Pontefract the slope of the mount must have been much
reduced before the walls of revetment were added: the sandstone upon
which the castle was built is soft, and would lend itself easily to
such an operation.
The bird’s-eye view of Pontefract just mentioned cannot be regarded
as absolutely trustworthy, but it gives us the relative position of
the various towers of the castle. It shows us a curtain flanked by a
formidable row of mural towers; the keep, a complicated erection of
several segments, with bartizans[223] projecting from the battlements
in the angles formed by the junction of the segments, is still the
dominant feature of the castle; but our attention is equally claimed
by the defences of the curtain and the domestic buildings which it
encloses. And, in pursuing our subject, we must first trace the growth
of domestic buildings within the castle area, and then turn to that
strengthening of the curtain which led eventually to the disuse of the
keep.