[image ALT: Much of my site will be useless to you if you've got the images turned off!]
Bill Thayer

[image ALT: Cliccare qui per una pagina ditch aiuto in Italiano.]

[Link to a series of help pages]
[Link to the next level up]
[Link to my homepage]

[image ALT: link to previous section]
Chapter 2
This webpage reproduces a chapter of
Romano-British Buildings and Earthworks

by John Ward

published by Methuen & Co. Ltd.
36 Essex Street W. C., London

The text, and illustrations except as noted,
are in the public domain.

[image ALT: link to next section]
Chapter 4

p47 Chapter III


Their Fortifications

Ramparts and Ditches

The earliest fortifications were probably mere banks of loose stones gathered from the surface. But in all ages since man learned to dig, the ditch with its correlated bank of upcast has been recognized as the most effective defensive line with the least expenditure of labour, as the ditch itself constitutes an obstacle, and its depth accentuates the relief of the slope which confronts the assailant. But the slope of a bank of earth is necessarily low, and to offer further impediment ditch and slope may bristle with stakes and other obstacles. Sooner or later the advantages of a steeper slope would be recognized. Various methods of attaining this would suggest themselves, as a facing-wall of stone or turves, a facing-row of posts driven into the ground, or the introduction of bonding courses of logs or brushwood in the bank itself. Gradually it would be realized that a rampart need not consist of the upcast from a ditch at all, but may be wholly constructed of other material. In the Antonine Wall, for instance, the soil from the ditch was disposed along its front edge to form a glacis-like bank or spread, while the rampart was built of turves. And in some of the bastioned forts the ditch apparently was dispensed with, the massive stone wall alone separating assailant from defender.

Whether the above represents exactly the successive developments in the art of fortification or no, the whole gamut of transitions had already been passed through before the Romans set foot in Britain, so that an attempt to make rampart-construction a test of relative age in Britain seems likely to be futile; all that p48can be said with reasonable certainty is that the earth-ramparts were typologically earlier than those of built stone. We have boats of wood and iron, but because the latter are of modern introduction we do notº argue that every wooden boat is necessarily older than every iron boat, for these boats are still built, and this shows that wood construction has still advantages over that of iron under certain conditions.

We may dismiss, therefore, any idea that the engineers of the Scottish 'earth' forts — Ardoch, for instance — constructed their ramparts of earth because they knew no better; on the contrary, the arrangement and intricacy of these great works prove them to have been masters of their art. The clue undoubtedly lies in the multiplicity of their ditches. For reasons not clear to us, there must have been a local need for this multiplicity: perhaps the northern tribes were bolder and more aggressive than the southern. But granting these ditches, the immense volume of their upcast had to be disposed of somehow, and how better and more economically than by utilizing it for a rampart, strong in its hugeness? The single ditch of many a southern fort could only have provided material for a small bank, hence the desirability that its effectiveness should be augmented by steepness of face and the disposal of the earth to the best advantage. Both were attained at Gellygaer by a retaining-wall, which not only provided a vertical face and allowed of the upcast which otherwise would have been required for an outer slope being utilized to raise the bank behind, but supplied in addition more soil for this purpose from its own foundation trench. Ramparts of this type seem to have been frequent, and there is reason to think — as will be pointed out presently — that in some cases the earth backings have been removed or spread out. The question of the contemporaneity of wall and bank in these ramparts is an interesting one. At Gellygaer the wall can hardly be otherwise than part of the original design; but at Caerwent it was built long after the bank. Probably in every case the bank served as a rampart, if only for a short interval, as the first consideration would be to provide a defensive line as speedily as possible, and this, of course, would at once be supplied by the upcast of the ditch. Caerwent suggests that many of our 'stone' forts may have been originally designed as 'earth' forts. We have already remarked the apparent absence of p49outer ditch and inner bank from some of the bastioned forts, but whether they really lacked them can only be disclosed by more thorough exploration. Cardiff, however, has a bank, and although of comparatively small dimensions it is too large to be accounted for by the upcast of the foundation-trench of the wall, and thus suggests a ditch; but it may be in part the legacy of a fort of earlier type on the site. So far as appearances go, it would seem that if the more developed bastioned forts had ditches, that as time went on, more and more reliance was placed in the wall, which consequently became thicker, loftier, and more strongly constructed, besides increasing in efficiency by the addition of bastions?

We shall now dissect a few examples of Roman ramparts and ditches, beginning with those of Gellygaer,1 as they furnish an excellent insight into the methods of the Roman engineer. The ditch had the usual angulated or V‑section, approximately 19 ft. and 7 ft. deep. The rampart was set back from its inner lip about 5 ft., and its average width was a trifle under 20 ft. In addition to the facing-wall there were the remains of a thinner wall at the back of the earthwork to support the foot of its slope, and both were built of the local Pennant-grit, and rested upon foundations of rough pieces of the same, laid horizontally in trenches from 1 ft. 6 ins. to 2 ft. in depth. The front wall varied from 3 ft. to 4 ft. 3 ins. in thickness, and was more carefully constructed than the back wall; and it also remained to a greater height, which, however, rarely exceeded 3 ft. The facing-stones of both were in regular courses, but those of the front wall were larger, and were here and there slightly dressed by hammer, chisel or punch. The inner sides of these walls were extremely rough, showing that they had been built against the earthwork, and this and some other circumstances afforded an insight into the procedure of the builders.

First the ditch was cut, and its soil was thrown up to form a bank, leaving four openings for the gates. Then, after a longer or shorter interval, this bank was cut back to an upright face, and the foundation-trench for the front wall was dug, the soil from both operations going to augment the bank. Then followed the masons with their wall, which was returned 5 or 6 ft. at the p50corners of the entrances. Meanwhile, but certainly after the wall was carried up to some height, the bank was removed at the spots to be occupied by the guard-rooms of the gates and the turrets, the soil from these going to still further raise the remaining portion of the earthwork. This may seem a roundabout process, but it was clearly proved by the fact that the outer wall had been built against earth at these places as elsewhere. Then followed the stretches of inner retaining-wall, the ends of which abutted against the turret basements and slightly overlapped the guard-rooms, these ends being finished off as steps by which access was gained to the rampart-walk. This walk would be on the summit of the bank, for the wall was too narrow to have provided one. The wall would be carried up sufficiently high to form a parapet, and the back of the bank would slope down to the slighter wall behind, the whole having the section indicated in Fig. 15. The earthwork was about 14 ft. wide. Assuming that all the upcast from the ditch and the foundation-trenches of the front wall and other structures described above, besides that from the sites of the guard-rooms and turrets, contributed to it, the height of this earthwork would be about 11 ft. Assuming this height, that of the wall to the top of the merlons could not have been less than 16 ft.

[image ALT: zzz]
Fig. 15. — Section of Ditch and Rampart, Gellygaer, (restoration of rampart in broken lines).
(15 ft. to 1 in.)

The wall at Housesteads2 was somewhat thicker than that at Gellygaer, but was faced on both sides. Remains of a bank behind it were found in 1898, as also remains of an inner revetment, showing that the total thickness of the rampart was from 24 to 26 ft. The Rev. J. Hodgson observed it as "a terrace, made of earth and clay, which ran from turret to turret along the inside of the wall to the height of 5 ft." Possibly this earthwork was never high enough to have provided a rampart walk; but the fact that late Roman buildings were found erected against the inside of the wall and on the level of the interior of the fort, may indicate that at their period the bank was not considered to be of use, and so was removed to a considerable extent. This seems to have also been the case at Great Chesters,3 where the remains of buildings, all apparently of late work, have been found in a similar position. At the east end of several barracks at Chesters (p100), the intervallum was laid bare some years ago and between this and the rampart-wall was a vacant p51space about 12 ft. wide, which apparently represents the bottom of a bank, the remains of which were probably removed unnoticed in the course of the excavation.

The wall of Caerwent4 contrasts with the foregoing examples in its great size and better preservation, standing in places to a height of 20 ft. or more. The face is vertical, rising from a projecting plinth of large tabular stones. The thickness at the base is from 10 to 11 ft., and this is reduced upwards by offsets at the back to 6 ft. 6 ins. at the existing summit. The foundation is about 3 ft. deep, and consists of rough stones laid without mortar, clay, or other binding material. The front facing is of the local limestone and sandstone, roughly squared, and laid in regular courses. The back facing varies considerably and is sometimes very rough, and this, together with the different heights and widths of the steppings, indicates that the wall was built in lengths by different gangs of men. The method of construction is clear. As each course of the front and back facings was laid, the intervening space was levelled up with rough pieces of stone, inclined or on end, and over their upper surface was spread gravelly mortar, but with no attempt to fill the interspaces. The wall has broad pilaster-like projections at intervals on the inner side, and attached to the exterior of the south wall are several large bastions. On the north side of the town two ditches have been cut through, both having a rounded V‑form, the inner being about 15 ft. from the wall and 20 ft. wide, and the outer somewhat smaller; while on the east and west sides a p52broad hollow marks the line of at least one large ditch. The earth-bank behind the wall is still an imposing feature, and, as already stated, the wall was a late addition.

The Roman wall of Cardiff Castle5 was brought to light a few years ago in the condition in which it was buried under the great medieval earthwork, and remaining to a height of from 12 to 15 ft. It resembles the wall of Caerwent in its general form and dimensions, but is more strongly and solidly built. The facings are of squared and slightly hammer-dressed blocks of lias limestone laid in mortar, the intervening space being packed with river boulders and broken limestone, all consolidated into a singularly hard mass with grout. The wall rests upon a foundation of these boulders deposited in a trench about 15 ft. wide and 1 ft. 10 ins. deep. Over these was spread a layer of mortar, upon which was laid the plinth-course, 11 ft. wide and 8 ins. thick. The wall above for the first 7 ft. 6 ins. is 10 ft. thick, but at that height it is reduced by four offsets on the inner side to 8 ft. 6 ins. The front facing rarely remains to a greater height than 3 or 4 ft., and its weathered condition tells of long exposure. The back facing, although of inferior workmanship, is better preserved and remains to the height of 10 or 11 ft. This is of that presence of a Roman bank of about this height, the soil of which was evidently derived from the foundation trench of the wall and a ditch, of which, however, nothing remains, as along its line the medieval castle-builders cut their largest ditch.

The wall of Burgh Castle6 also resembles that of Caerwent, but it presents a very different appearance, being constructed of flints, with tile courses at intervals. It still remains to a general height of 15 to 16 ft. The facings are of split flints with their fractured surfaces outwards. The flints are laid in mortar, pink with pounded brick, and the triple tile-bands occur at every five or six of their courses. The tiles are, as a rule, only one row deep, showing that their function is to lace the flint facings to the core. The wall is stepped back behind, reducing its thickness to 8 ft., and the facing on that side is rough and without tile courses. The foundation is said to be of puddle clay overlaid with a stratum of flints, in a shallow trench. It is supposed that the wall was lined with an earth-bank which has been ploughed down, but there is no evidence for this, nor for a ditch.

p53 The walls of Richborough, Lympne, and Porchester resemble the last two, except in their greater thicknesses. They are all strongly built structures of grouted rubble and boulders, with tile bands, and faced, at least on their fronts, with squared stones of local origin. The great thickness (14 ft.) at Lympne is due to the shifty nature of the soil, which demanded a wide base. According to Mr. Roach Smith, it simply rests on the natural surface; but it is probable, as in the case of the sea-wall at Burgh Castle, that it had a foundation of piles which have disappeared. Recently the walls of Pevensey, 12 ft. 3 ins. thick, have been found to rest upon piles. In none of these examples has a mound or a ditch been observed.

We now pass to the remarkable Scottish forts described on pp26‑30. The ramparts of Camelon, Lyne, Ardoch, and Birrens7 so closely resemble one another that we shall treat them together: but it may be mentioned that the last has several points of difference from the others. Although 'earth' forts, the structure of their ramparts is more complex than the 'tumultuary' work of Vegetius. Their defensive lines have been cut through in several places; but, as might be expected, the rampart sections were confused through the spread of the materials beyond their original limits. The sections (Fig. 16) are selected from those of the reports, but have been simplified to render comparison easy. These sections are as follows: The south defences of Camelon, and the inner portions of those on the east sides of Ardoch and Lyne and of the north side of Birrens, the space not admitting of the whole of the complex defences of these being shown.


[image ALT: zzz]
Fig. 16. — Sections of the Fortifications at Lyne, Camelon, Birrens, and Ardoch.
(20 ft. to 1 in.)

Underlying all these ramparts are rough pavement-like foundations (shown in solid black) laid on the old natural surface. At Camelon, we have an outer spread of rough stones, bedded in clay and varying in width from a single stone to about 4  ft. 9 ins., where excavated; and an inner, which also varied in width, and in one place consisted of two courses of stones. The outer was only found on those sides where there is a ditch. These strips are the footings of the inner and outer faces of the rampart, and indicate a width of about 40 ft. for the latter. The corresponding strips at Lyne are each about 4 ft. wide and indicate a rampart of about 23 or 24 ft. At Ardoch, the outer strip p54is much wider, being from 7 to 8 ft. in width, with roughly dressed stones along the outer margin; and the inner varies considerably, taking the form of paving, or a rough spread of cobbles, or a wall of several courses, the original width of the rampart being about 33 ft. At Birrens, the main underwork is from 18 to 19½ ft. wide, and is constructed of tabular polygonal stones neatly fitted together; and set back about 1 ft. from its outer edge was found in most places a narrow second course about 3 ft. wide. This pavement-like structure carried the main body of the rampart; but along the west side of the fort is a strong kerb, formed of a double row of large stones set in the ground, about 10 ft. from the inner edge of that foundation, which appears to have a bordered terrace behind the rampart.

The ramparts of these forts are of stratified structure. In that of Camelon, the base is of large pieces of split wood and branches mostly laid in a longitudinal direction and mixed with peat and clay. Above these were usually noticed thinner layers of the same; and still higher, consolidated sand and gravel with a little clay. In all sections was observed a mass of puddled clay resting on the outer stonework, and tailing outwards above. There is no doubt that this clay was the facing of the rampart, and that the tailing was due to the spreading outwards of its upper portion. Clay in a similar position at the back of the rampart was observed here and there. The rampart at Lyne, which is much worn down, is formed of layers of clay and black mould; that of Ardoch, of layers of clay, gravel, turf, and brushwood, with masses of clay on or about the footings, tailing outwards as at Camelon; and that of Birrens, of earth, clay, sods, and brushwood, also in layers; but the remains of clay facings were either absent or not reported. The split trunks and branches at the base of the Camelon rampart may have formed a rumble drain, to keep the earthy materials above dry; if so, the clay and peaty matter, "wet and slimy," that filled the interstices, were probably washed down from the layer above. Although no mention is made of outlets through the outer footing to carry off the accumulated water as in the Antonine Wall, they might easily have escaped notice, as the exploratory cross-trenches were few. The presence of timber is not mentioned in the reports of the other forts, but the brushwood observed in the ramparts of Ardoch and Birrens may have served a similar purpose.

p56 The ditches of these forts, with the exception of those of Birrens, are of the usual angulated form, but those of Ardoch have rounded bottoms, due, perhaps, to water-wear. The form at Birrens is peculiar in having convex sides and a flat bottom. With the exception of the small inner or berm ditches at Camelon and Ardoch, they range from 11 or 12 ft. in width at Lyne, to 23 ft. at Camelon, and in depth from 5½ ft. to 8 or 9 ft. The small ditches just referred to are less than half the size of their neighbours, and their function appears to have been to keep their respective ramparts dry. The excavations at Camelon proved that in one place at least the berm ditch contained split timbers like those at the bottom of the rampart; and in two cuttings at Ardoch cobble stones were found at the bottom, "which appeared to have been thrown in to form a drain."

The remains of the ravelin ramparts or parapets and outer banks at Ardoch and Lyne, and of the rampart of the great annexe or 'south camp' at Camelon, show no laminated structure and appear to consist of the upcast from the ditches. They also lack the stone bottomings of the main ramparts. The defences of this annexe have an interesting feature. The interval or platform between the innermost and second ditches is expanded on the south and east sides to 27 ft. in width, and a little behind its inner line is a small V‑shaped trench, 3 ft. wide. It is too small to have been a defensive obstacle, and from its position it was useless as a drain. It has been suggested that it held a palisade.

The rampart of the free sides of Rough Castle, like the Antonine Wall which closed in the fort on the remaining side, is of turfwork. It rests upon a stone bottoming that varied considerably, "of an average width of not less than 20 ft., supplemented by varying margins adapted to suit special requirements, and increasing the width so that it was nowhere less than 30 ft." The rampart appears to have been originally about 20 ft. in width, but was afterwards supplemented by additions on both sides, making a total width of about 34 ft. The ditches are, as usual, V‑shaped, about 16 ft. wide and 8 ft. deep. They are separated by a narrow strip of the original surface, which is capped with a layer of firmly bedded stones, as also are the tops of the opposite brinks; in fact, "this stone lining, at all parts liable to be easily damaged, is a noted characteristic of the whole work." The upcast of the p57ditches was used here and there for a glacis-like mound and other external works. Of the rampart of the annexe little remains, but the absence of lamination indicates that it was a simple earth one. It also rests upon a spread of stones, which, however, extends inwards to serve as a roadway 15 ft. wide behind the rampart. The annexe ditch is similar to those of the forts, and its brinks are similarly capped with stonework.

The excavations at Coelbren brought to light several remarkable features in the construction of the rampart. Along the south side where the ground was treacherous, and at the four angles, the rampart was raised upon a bottoming of large oak logs. They were laid transversely in a shallow trench, and were nearly 18 ft. long, or two short logs were used to make up that length, the whole forming a sort of rough corduroy. Upon this foundation was about a foot of dark soil containing decomposed vegetable matter, presumably derived from the trench below. Then followed a layer of branches, mostly of birch, laid irregularly and loosely, perhaps to serve as a bonding-course. The remaining portion of the rampart above was of turfwork. There was evidence that the scarp had been faced with white clay; while in most of the cuttings, the rear portion of the rampart was darker than elsewhere and extended several feet behind the logs, indicating apparently a greater width for the rampart than the corduroy.

At the rounded corners the logs were laid fan-like at right-angles to the curve, and these platforms were pinned down by stakes driven into the soil beneath. At the south-east corner there were two platforms, one above the other. The cuttings on the remaining three sides of the fort yielded no trace of a log foundation; but in one or two, the rampart rested on a layer of brushwood. The rampart itself also varied, the turves occasionally appearing to be mixed with earth or clay. Colonel Morgan, in remarking the careful and strong construction of the log foundations of the corners, came to the conclusion that "the engines for missiles were placed only on the angles, as they alone would have necessitated this unusual foundation." The berm on one side at least, was 16 ft. wide. The two ditches are described as V‑shaped, the width of the inner averaging 9  ft., and the outer, 7 ft., but the exact depths were not ascertained. There were clear indications of a glacis-like spread of clay, doubtlessly derived from one or both ditches. In the ditches were found a p58number of "oak stakes, 9 to 12 ins. long, pointed at one end, with a curious notch below the point, which the Colonel considered to be portions of obstacles placed on the berm."8

Turrets and Bastions

Classical writers mention turrets in connection with fortifications. They are represented on the Column of Trajan, and two of these are shown in Fig. 17, the one set diagonally within the rounded corner of a fort, roofed, and apparently constructed of wood as no masonry joints are shown; the other, certainly of stone, with a flat top. In the Naples Museum is a remarkable bronze water-heater from Herculaneum, in the form of a small embattled fort with an embattled tower at each corner. The corners of a rectangular fort are its most vulnerable points, as only a few defenders can there be accommodated, and they are liable to be opposed by a large number of assailants; hence the value of turrets in these positions, as they increase the accommodation of the defenders and give them greater 'command.' It is precisely in these positions that the structures we are considering are most usually found. They are shown there, and only there, on the plans of Melandra, Brough, Hardknott, Great Chesters, and Castlecary; while on those of others, as Gellygaer, p59Housesteads, and Chesters, there are in addition similar structures along the sides. The fallen débris that choked the east corner turret at Gellygaer,9 contained much wood-charcoal (indicating a destruction by fire) and broken roofing tiles. The latter were near the top, while the former lay at various levels below, showing that the structure was roofed with tiles and that there was much timber-work, some presumably relating to floors, below the roof. More definitely the excavation of the Mucklebank turret10 on the Wall of Hadrian, described on page 124, proved the former existence of an upper floor.

[image ALT: zzz]
Fig. 17. — Fortification Turrets from Trajan's Column

At Silchester and Caerwent the walls have at intervals broad pilaster-like projections on their inner sides, but whether they should be regarded as the bases of turrets is by no means certain. At the former11 they are about 12 ft. wide, and at the latter,12 somewhat wider; and in both the projection is from 18 ins. to 2 ft. They are really portions of the wall carried up the full thickness of its foot; but the one within the south-east corner at Caerwent is an exception, having a greater projection than usual. These 'counterforts,' as they are sometimes called, were certainly not buttresses, as the walls of these cities were too thick and strong to require such supports.

Perhaps High Rochester throws a light on their use. Here the wall for about 80 ft. to the south of the west gate, and nearly 50 ft. to the north, is increased on its inner side to a thickness of nearly 30 ft.; and on the south wall a shorter length is similarly thicker than elsewhere.13 It is supposed that these thicker portions provided platforms for the great engines for hurling stones; and the existence of such platforms is proved by the discovery of a tablet near the west wall recording the construction of a ballistarium in the time of Caracalla, and of another inside the fort recording the restoration of one. Moreover, in this, as in several of the Wall forts, many large rounded stones weighing a hundredweight or more have been found, which certainly were the missiles of ballistae. Some thickenings of late work on the inner side of the wall of Housesteads may have had a similar purpose. As the 'counter-forts' at Silchester and p60Caerwent increased the summit of their respective walls to about 10 ft. in width, they would provide solid platforms for military engines of considerable size.

It is noteworthy that no traces of turrets have been reported in the case of the Scottish 'earth' forts. This, however, does not disprove their former existence, as they may have been of timber. At Coelbren the corners of the fort rested upon specially strong foundations (p57); but this may indicate nothing more than that it is just at these points where a rampart requires special strength. The patches of stone foundation at Castleshaw, on the other hand, are not under but behind the rampart at these points; but whether they supported turrets or ballistaria is uncertain.

Bastions have already so frequently been referred to that their forms need not detain us further, beyond a reference to Fig. 18, which gives their plans to a common scale. They are normally solid structures, at least to their existing heights; but there are several exceptions. At Cardiff, the middle bastion14 of the east side is solid to the height of 6 ft. 6 ins.; but above that height it was found to enclose a chamber of its own shape. In our next section reasons will be given for thinking that this chamber contained a postern (p33). At Caerwent are the remains of three large polygonal bastions along the western half of the south wall. Each is solid below and has a chamber above with a mortar floor, and at intervals on its level drain-holes through the outer walls, several of which retain a semicircular channel of mortar. These bastions are peculiar in another respect. They are not parts of the original construction of the town wall. Their foundations are separate and deeper; and when they were built large holes were roughly cut in the face of the wall, into which their masonry was toothed. Portions of their outer walls still stand 11 ft. above the internal floors, but without any signs of loopholes or other openings.15 It would appear from this that their basements were not used for defensive purposes, but probably for storage; and they must have been reached from above, as the rampart-wall is continued along the back without a break.


[image ALT: zzz]
Fig. 18. — Plans of Turrets and Bastions.
(30 ft. to 1 in.)

At Cardiff, the bastions are bonded into the wall and are of one construction with it, as also at Lympne and Pevensey. At p62Richborough and Burgh Castle, on the other hand, the lower portions of the bastions are built against their respective walls, but their upper portions are bonded into them. This has given rise to the supposition that these forts were originally without bastions, and were afterwards ruined or partly pulled down and reconstructed on bastioned lines; and a well-defined break in the core of the Burgh Castle wall, about 8 ft. above the ground, is held to substantiate this. But the writer finds that the summit of the lower part of the wall at this break is not rough, like an old ruined wall, but is finished off, roof-like, and smoothed over with mortar, as if with the view to prevent the access of rain to the core below.16 Externally, the whole work has every sign of being the production of the same builders, the facework above and below being identical in appearance. The break seems to represent a halt, conceivably a winter's cessation of the building operations; and the omission of the bastions in the first stage of the work may simply be due to a desire to raise a barrier with the least expenditure of labour before the winter set in. At Richborough, the corners of the lower portion of the wall are rectangular, and if they represent an early unbastioned work, we have the anomaly of a return to an old type of Roman fort that was abandoned before the conquest of Britain, for one with rounded corners, and this certainly militates against the theory that the bastions were an afterthought.

With regard to the original heights of the bastions, their remains at Cardiff, Caerwent, Burgh Castle, and Pevensey indicate that they were at least as high as the existing remains of the walls. At Burgh Castle, long stretches of the wall have the uniform height of from 16 to 17 ft., and the flat tops of the better preserved bastions are as high, giving the impression that this level approximates to the original height of the whole work, less the parapets. It is possible, of course, that the bastions were surmounted with structures of timber or of sailor masonry; but their tops have the curious feature of a central shallow hole about 2 ft. in diameter, which may have received the pivot on which a military engine revolved.16

p64 Gates and Posterns

With few exceptions to be noticed presently, the gates of the forts and fortified towns of which we have any knowledge, were stone structures. The masonry was usually better and more massive than that of the ramparts; and as they received some degree of architectural treatment and embellishment, they must, in a pleasing manner, have broken their monotonous lines. The tablets recording their construction or restoration were sometimes highly ornamented, as one found at Risingham and another at Lanchester indicate.17 The structures varied greatly, but a considerable number followed a common model, and examples of those will be considered first.

The north and south gates of Caerwent18 are excellent examples of gates with single passages. They are of like size and design, and while the south gate is the better preserved, the north gate still retains portions of its external front, which has fallen in the other. In each, the general structure is apparently older than the rampart wall, and is rectangular, about 15 ft. wide and 14 ft. deep, with a passage 9 ft. 6 ins. wide. The passage is contracted at the front and back to 8 ft. 9 ins. by projecting jambs. These had moulded imposts and carried arches, portions of which remain. Fig. 19 is an elevation of the back or town front of the south gate, with the wall abutting against the sides of the structure: the external front was probably similar. In the angles behind the front jambs of the north gate are still to be seen the blocks of stone, level with the roadway, which contain the sockets in which the door-pivots turned. The doors were in two leaves, which, when open, fell back into the recesses between the front and back jambs.


[image ALT: zzz]
Fig. 19. — South Gate, Caerwent. Elevation of inner face, with inner profile of Town Wall on either side.
(4 ft. to 1 in.)

The architectural treatment of the two gates is conjectural. The rampart-walk was somewhat higher than the crown of the arches, and was probably continued over the space between them by a timber floor. Two pieces of moulding found near the south gate may have belonged to a cornice above the arches; and the many roofing-tiles about the site suggest that the passage of the rampart-walk was through a covered chamber, p65as in the third and fourth illustrations of Fig. 20, which are fortification gates sculptured on the Column of Trajan. The first and second lack upper chambers, and in the latter is shown the timber parapet of the rampart-walk over the gate. The second two have chambers with windows over the portals, and doorways at the sides by which they were entered from the rampart. The arched entrance and windows of the fourth example show that it was intended to represent a stone structure.

[image ALT: zzz]
Fig. 20. — Fortification Turrets from Trajan's Column

Of similar character were the gates of the mile-castles of the Wall,19 of which each had two, level with the fronts of their respective walls, but projecting behind. Those of the mile-castle near Housesteads are the best preserved. The widths of the openings are nearly 10 ft. The massive jambs have plain square caps on which still remain the springers of the arches, each gate having two as at Caerwent.

The remains of two-passage or double gates may be seen p66at Housesteads, Birdoswald, and Great Chesters; but those at Gellygaer20 indicate a somewhat simpler construction. All the gates of this fort were precisely alike, but the south-west one was the most thoroughly explored. The passages of this gate, as will be noted in Fig. 21, were similar to those at Caerwent. Their contracted openings were also of similar width, and the pilasters were arched, as indicated by the well-shaped voussoirs of calcareous tufa found about the sites. One of the thresholds still remained intact and consisted of two long flagstones containing the sockets for the door-pivots and two square bolt-holes, with a raised rim on the outer side formed of two other flagstones set on edge in the ground. This rim sheathed the bottom of the doors when closed, and it exhibited two worn hollows about 4½ ft. apart, made by the passage of wheeled vehicles. On either side of the gate was a guard-chamber, the front of which was a continuation of the rampart-wall, and in the back was the doorway by which it was entered. The front of the gate was set back from the rampart face nearly 6 ft. That these gates, or some portions of them, were roofed with tiles, was proved by the broken red roofing-tiles about their sites.


[image ALT: zzz]
Fig. 21. — Plan of South-west Gate and Ditch with sections, Gellyager
(15 ft. to 1 in.)

The double gates of the Wall forts mentioned above were similar, but of stronger construction, and this is especially noticeable at Housesteads. Those of Birdoswald21 most closely resembled the Gellygaer gates in their planning, but were on a larger scale, the openings being nearly 12 ft. wide, Fig. 22. Those of Housesteads, as also of Chesters and Great Chesters,22 differed in two respects. The intervening wall between the passages, instead of being solid, had a central opening, probably arched; and the guard-chambers were entered from the passages. All the gates were set back from their respective rampart faces, but those at Housesteads less so than the others; while the guard-chambers at Great Chesters had the unusual feature of being slightly in advance of that line. The thresholds were generally constructed of a row of large stones, with their outer margins raised to form a rim, and in a few instances there was a central stone stop-post as well. The door-socket were sometimes cut in the bottom stones of the jambs which projected for the purpose.


[image ALT: zzz]
Fig. 22. — Plans of Roman Fort and Town Gates
(30 ft. to 1 in.)

p68 While the castella named above had double gates, the supplementary gates at Birdoswald and Chesters were single ones. The north and south gates at Caerwent were not the principal ones, which may prove upon excavation to have been double. Small forts, like the mile-castles, usually had single gates, but there were large forts with only single gates. High Rochester is an example.23 To judge from its west gate, which is the least ruinous, they were of simpler character than those of Caerwent, having a single pair of jambs each, deeply set back from the rampart face. The lateral gates at Birrens24 appear to have closely resembled those at Caerwent, and the north gate was remarkable for the great length and narrowness of its passage. Those at Castlecary25 were very ruinous, but were apparently of a single span each. Those at Camelon26 were still more ruinous. Their side walls were from 20 to 22 ft. apart, but on either side of the roadways were deep post-holes which reduced the width to the proportions of a single span. In three of the gates at Bar Hill27 were found the stumps of oak posts in like positions, three on each side. According to Dr. Macdonald, these posts retained the vertical ends of the turfwork rampart and supported timber gangways, but were not the posts of the actual gates; but it is difficult to understand why they should not have fulfilled all three purposes. The Lyne gates28 were wholly of timber, simple, and of a single span each. At Ardoch29 they were also of timber, and the post-holes of the east gate indicated a complicated structure of the depth of the rampart and divided into three parallel spaces, of which the middle was apparently the passage, and the outer possibly guard-chambers. It will be noticed that this gate alone of the examples given in this paragraph had traces of these chambers.

On the sites of some of the gates of the Wall forts, and notably at Birdoswald, have been found door- and possibly window-heads. Similar heads were used in Norman and Early English work, but the Roman examples, when otherwise than plain, are treated as sunk panels containing ornamentation in the spandrels. They are generally regarded at heads of the guard-chamber doors, p70but at Birdoswald they are numerous, and some are rather small for doorways. These may be window-heads, and may have belonged to the windows of anº upper structure.

The walls of the guard-chambers are usually of considerable thickness, as though to sustain lofty superstructures; and the resemblance of these chambers to the turret basements is decidedly convincing in this respect. It is interesting to find that a gate figured on a mosaic in the Avignon Museum has its guard-chambers carried up as two turrets. In Fig. 23, this gate is reproduced from Collectanea Antiqua.30 It has two arched portals with three windows above, and on either side will be noticed the window of a guard-chamber and two smaller ones over it. The whole structure, as also the rampart-wall, is embattled, and the merlons of the latter and of the turrets are wide apart and have projecting copings as in the sculptures of Trajan's Column, while those of the middle portion of the gate are closer together and are not capped. The delineation admirably fits in with what we know of the double gates in this country, but the roofs are shown as flat to accommodate defenders, whereas at Caerwent and Gellygaer there is evidence for tiled roofs. A glance at Figs. 17 and 20 will show that the Romans did not exclusively adopt one or the other, but it is probable that in a rainy country like ours gates often had tiled roofs.

[image ALT: zzz]
Fig. 23. — Gate of Fort on Mosaic, Avignon Museum

The gates to follow not only differ from those already described, but they differ more or less from one another. The Balkerne or west gate of Colchester appears to have been on an unusually p71imposing scale, and its ruins are one of the most conspicuous vestiges of the Roman town. Like the wall, it is built of a local chalkstone with lacing-courses of tiles. The southern third of the structure remains to a considerable height, and consists of a narrow arched passage and a quadrant-shaped guard-chamber. The northern two-thirds, with the exception of the curved wall of the other guard-chamber, have long been removed, and the site is occupied by an old inn. Mr. Roach Smith31 and Dr. P. M. Duncan32 respectively, described the remains in 1847 and 1855, and both considered that they indicated a gate with a wide carriage way and a narrow one for foot passengers, or possibly two, one on either side of the former, with a guard-chamber to the south, and one or two larger chambers to the north. The writer, however, suspected that the structure was symmetrical, with two carriage ways and two for foot passengers, the whole being flanked by two quadrant-shaped guard-chambers. An examination of the remains somewhat confirmed this, and it was further confirmed by measurements and a plan made by Mr. Arthur G. Wright, the curator of the Colchester Museum; but without the evidence of the spade it is hardly possible to go further. In the plan, Fig. 24, the visible remains are indicated in black. It will be observed that the whole structure is in advance of the town wall, also that the outer or curved walls of the guard-chambers are thinner than the intervening walls. This is suggestive that the main fabric of the gate was rectangular, 60 ft. in width and 30 ft. in depth, and loftier than these chambers, with two large arched ways flanked with two smaller, and a storey above, the p72whole probably resembling the Porte d'Arroux at Autun and having a similar series of arched openings above the portals.

[image ALT: zzz]
Fig. 24. — Plan of the Balkerne Gate, Colchester
(30 ft. to 1 in.)

A considerable portion of the north gate of Lincoln — the Newport Arch — is standing, but is buried to the extent of about 8 ft. in the soil and débris accumulated since Roman times. The structure is about 34 ft. deep and has a single passage for the road, 17½ ft. wide. The inner or back portal of this passage is still intact, and is nearly 16 ft. in the clear and rises to a height of about 22½ ft. above the Roman level. Its arch is of a single ring of large limestone voussoirs rising from imposts which appear to have been moulded. The outer or front arch has long since disappeared. On the east side is a postern for pedestrians, 7 ft. wide and contracting to about 5 ft. at the north end, and 15 ft. high from the Roman level. On the west side there was a similar postern about a century ago. The whole structure is of good masonry, and it appears to have projected considerably beyond the north face of the town wall.

[image ALT: zzz]
Fig. 25. The 'Newport Arch' (Roman North Gate), Lincoln

The other gates of Lincoln33 appear to have been of like form, size and construction. The west gate is buried in the post-Roman earthwork of the castle, but its front was exhumed in 1836. The excavation was deep enough to expose the arch of the carriage-way, which was of precisely similar character to that of the Newport Arch; but the most interesting feature was the remains of the storey above. The weight of soil had considerably pushed its masonry out of the perpendicular, but enough was left to indicate that there were three window-like openings over the arch, and these are said to have been 4 ft. wide. In a contemporary lithographed view34 of this gate, one of these openings is shown remaining to the springs of its arch, apparently 5 ft. or more high and between 3 and 4 ft. wide. One side of the gate was sufficiently disclosed to show a similar opening on the same level. The excavation was not deep enough to reach the posterns; but the structure as shown in the view does not seem wide enough to have included these, and it is obviously out of scale, not agreeing with the few measurements given. A comparison of the two gates (the north and the west) indicates a frontage-width of about 47 or 48 ft., and an original height of not less than 40 ft. How the summit was treated p73is, of course, a matter of conjecture. The sides, like the fronts, may have had three openings each, the third deeper and serving as a doorway from the parapet-walk of the wall. Possibly the guard-chambers were external, flanking the ground storey of the main fabric, as seems to have been the case at Colchester.

The four principal gates of Silchester,35 of which the east and west were double ones and the north and south single, differed from all described above in having definite means of enfilading them. This was accomplished by their structures being deeply set back between incurved returns of the rampart-wall (see Fig. 22) hence an assailant would not only be resisted by the defenders of the gate itself, but would be subjected to the cross-fire from these returns. The north and south gates resembled the corresponding gates at Caerwent. The west gate was more complex. The returns of the rampart-wall were of great thickness, as if to serve as ballistaria, but more probably for another reason. The gate had the usual guard-chamber on either side, and in addition, a room entered from it, in south of the front of the gate and constructed against the returns, the two rooms forming an oblong structure with walls of considerable thickness. It is probable that these two structures were carried upwards as two towers, and that the great thickness of the returns was to provide suitable substructures for the outer side-walls of their advanced portions, and in addition, space for access from the rampart-walk to doors in the walls. It will be observed that the inner sides of these advanced portions of this west gate and the returns of the rampart at the north gate provided a considerable length of flanking defence for their respective gates. The east gate was similar but deeper, and if anything, of stronger construction.36 All these gates had brick arches, and timber thresholds with sockets near their ends to receive the door-pivots. The iron sheath, 3⅝ ins. in diameter, of one of the sockets was found; also two U‑shaped iron straps that were apparently used to bind the doors, and indicate for these a thickness of about 4 ins.

The north gate at Cardiff Castle, Fig. 22, has a single opening with an outer and an inner pair of jambs, the depth of the passage being 10 ft., representing the thickness of the wall; but the guard-chambers have projecting polygonal fronts like p74those of the bastions, only a trifle smaller. The original door-sockets were in backward projections of the bottom ends of the front jambs; but at a later date the roadway was raised and two large socket-stones were introduced at a higher level. These sockets have the unusual feature of a shallow recess cut in the side, which evidently received a corresponding projection on the iron lining or shoe, to prevent it revolving with the pivot. The later roadway, at least, had no ledge or rim across the threshold, but instead, a central stone door-stop.

The only gate at Richborough that has been explored37 was found to contain a single passage between two oblong guard-chambers which boldly projected on either side of the wall, Fig. 22. There were two pairs of jambs, and the outer were chambers set back in the passage. It is evident that the guard-chambers were an important feature, and were carried upwards as two large towers; and their bold projection would more effectually protect the approach to the arched opening than at Cardiff.

[image ALT: zzz]

The remaining gate at Porchester seems to have resembled that of Richborough. That at Burgh Castle is now a mere gap in the wall, but its width admits of a similar gate structure, also of a single span. The great gate at Pevensey resembles that at Cardiff in being flanked with bastions; but these bastions are precisely like the rest in this remarkable fort, solid, of great projection, and with rounded fronts. Of the gate-structure itself few traces remain; but it appears to have contained a single passage between two oblong guard-chambers as at Richborough. The whole, however, was so far set back between the bastions that the space between these formed a cul-de‑sac about 30  ft. wide and nearly as deep, thus providing accommodation for a large number of defenders.

Posterns are not found as parts of the original construction of the forts of the earlier type, the two additional gates at Chesters and Birdoswald being too large to be regarded as such. It is of common occurrence that double gates have been reduced to p75single openings by blocking up one of the portals, and that the remaining openings, as also those of single gates, have been curtailed to the proportions of posterns. In most cases these changes were effected in late Roman times. At Ribchester, Mr. Garstang discovered the remains of a curious sunk passage, 3 ft. wide, which passed through the turret and rampart-wall at the south corner of the fort; but as it made a right-angled turn and apparently ended in a well, it seems less likely to have been a postern than a passage to obtain water.38

Posterns, however, seem to have been usual in the bastioned forts. At Burgh Castle a small and simple opening through the wall has been observed close by the middle bastion on the north and south sides. At Richborough the middle bastion on the north side conceals an ingenious narrow passage slightly above the ground-level, which ends in a narrower portal in the east side of the bastion, Fig. 27. At Cardiff the middle bastion on the east side contains a small chamber, as stated on page 33. From the analogy of Richborough it is probable that this chamber was connected with a postern. The back of the chamber is continuous and remains to the height of the Roman wall generally, so that the access was probably from the rampart-walk by means of a ladder. Portions of the outer wall remain, but it is noteworthy that it is almost wholly broken away on the north, so that there may have been a small doorway on that side of the bastion from which the ground outside the fort was gained by another short ladder. It is probable that, as at Burgh Castle, the opposite sides of the last two forts had similar posterns. If it is permissible to draw a conclusion from very slender evidences, it is that these quadrangular bastioned forts still retained the four entrances of the Hyginan type, but that two of them, possibly representing the lateral gates of that type, were now reduced to mere posterns.

[image ALT: zzz]
Fig. 27. — Posterns at Pevensey and Richborough.
(30 ft. to 1 in.)

p76 Lympne and Pevensey differ from the foregoing in their irregular form and the unsymmetrical disposition of their entrances. In the east wall of the former may be seen two small gaps about 5 ft. wide, which were probably posterns. Pevensey had at least one postern. One was recently opened out in the north wall. Its passage is curiously curved and widens inwards, the internal orifice being about 8 ft. in width, and the external considerably less, Fig. 27.

The approaches to the gates were of two kinds — 'causeway' and 'bridge' approaches. In the former, the ditch or ditches were discontinued in front of the gate, leaving a space sufficiently wide for the road. In the latter the ditch was continuous, and the road crossed it on a bridge. The approach to the south-west gate at Gellygaer, Fig. 21, is a good example of the second kind. Immediately in front of the gate, the ditch was found to have its sides stepped out for about 18 ft. These steps undoubtedly received the supports of a wooden bridge. The inner half of each had a shallow chase which was filled with the remains of concrete. The earth immediately above showed signs that a beam had rested upon the concrete, and this apparently was the sleeper of the supports of the bridge-platform. Perhaps the middle and wider span of the platform was made to draw up, and was operated by chains from the upper part of the gate. A cutting in front of one of the smaller gates at Silchester39 proved that the ditch at that point was 80 ft. wide, this greater width than elsewhere being due to the suppression of the berm. Near the middle of the broad flat bottom was found a low gravel bank that evidently supported the trestles of the bridge, and it is possible that the shorter span, next the gate, was a drawbridge. A similar bank was found in the ditch, which was here 76 ft. wide, in front of the Roman gate at Aldersgate, London, about twenty years ago.40 The ditches are continued in front of the gates at Caerwent, and at Great Chesters, Birdoswald, Housesteads, and probably other Wall forts. The 'causeway' approaches were usually simple and direct, that is, they were at right-angles to the line of the rampart; but in some of the Scottish forts they were more or less devious or even intricate. At Lyne, that to the east gate was direct between the returns of the ditches, and was flanked at the p77entrance by the expanded ends of the rampart of the outer terrace. The opposite gate had a similar direct approach, which was dominated by an isolated mound or traverse at some distance from its entrance. The approaches to the lateral gates, on the other hand, had a somewhat zigzag course, instead of in front of the gate. The north gate at Ardoch was reached by a long oblique causeway through the intricate outer works on that side of the fort. The approach to the east gate is in good preservation, and is at right-angles to the rampart, but its entrance is rendered oblique by an angled extension of the outermost ditch. This approach yielded upon examination some interesting evidence of timber protective works. Along each side were a number of post-holes of a strong fence or palisade, and across it at intervals were others apparently of three gates. Remains of such structures have not been observed elsewhere: they were looked for at Lyne, but were not found.

Summary as to Sequence

While typologically the 'earth' forts may be older than the 'stone' forts, a strong doubt was expressed on page 47 whether this in itself could be accepted as a test of age in Britain. The rampart of Caerwent, it is true, was of earth only, before it was faced with wall, and recent investigations have proved that some, at least, of the Wall forts had originally ramparts of earth or of turves. But although the latter forts were remodelled, and even, in two or three instances, enlarged when reconstructed in stone, the new work was still on the lines of the camp of Hyginus — the lines also of the Scottish 'earth' forts. If Birrens had a stone wall and gates like Housesteads, Chesters, and Gellygaer, it would as little differ from them in this general planning as these do from one another, except in its numerous ditches. All we can say for certain, so far as present evidence goes, is that no 'earth' fort was raised during the second half of the Roman era in Britain.

If, on the other hand, the fortifications of the Wall forts are compared with those of the bastioned forts, we at once observe differences that can only be explained by a change in the principles of defence. In the one group we have walls that rarely exceed 6 ft. in thickness, internal turrets, and four large double gates: p78in the other, walls 9 to 10 ft. or more thick, external bastions, gates of a single opening each, and postern. The gates, moreover, of the former closely adhere to one model; while those of the latter not only differ from them, but show little agreement among themselves, and instead of four, there were two at most. The Wall forts are rectangular and symmetrical: the bastioned forts differ greatly in shape, several being rectangular or approximately so, while Lympne is an irregular pentagon, and Pevensey somewhat oval.

These modifications in the bastioned forts had a twofold effect: they increased the passive resistance against attack by their greater strength of structure and the restriction of the entrances, and they increased the active resistance by providing means of enfilading both walls and gates. That they indicate a difference of period cannot be doubted.

The Wall forts had been reconstructed in stone by the time of Caracalla, and the bastioned forts of the Saxon shore are a legacy of a later time, when that shore was threatened by oversea enemies. According to the Notitia, both series were held by garrisons at the close of the fourth century, and the abundance of late coins found on their sites corroborates this, while the numerous alterations seen in the Wall forts bear witness to their long occupation at the time they were abandoned. The nature of some of these alterations is significant. The curtailment of gates by late masonry, converting double gates into single ones, and reducing the widths of some of the remaining entrances — and even the complete walling up of other gates — were apparently in response to the same conditions which gave the bastioned forts their limited gate accommodation, both in number and size.

Gellygaer and Cardiff, so near one another, well illustrate what has been said above.41 The one presents a singularly perfect plan on the Hyginan model: the other was a bastioned fort. At Gellygaer there was no trace of alterations or other signs of a long occupancy, and its coins stopped short with Hadrian. Cardiff, although not mentioned in the Notitia, was apparently garrisoned to a late date, as the site yielded coins of Carausius, Constantine and Julian the Apostate.

p79 Caerwent and Silchester stand somewhat apart. Their walls are of the form and massive construction of those of Cardiff and Burgh Castle, yet without bastions — those at Caerwent being subsequent additions. Instead of turrets of the type of Gellygaer and Housesteads, they have solid pilaster-like structures. In both, the ramparts were originally of earthwork only. The remaining gates at Caerwent are of the earlier type, but they are older than the wall. Those of Silchester are apparently coeval with the wall, and they are of intermediate character, two being double gates, and all being provided with flanking defences, but by a different method from those of the bastioned forts. Perhaps we shall not be far wrong if we assign the forts of the type of Gellygaer, Housesteads, and Birrens to the first and second centuries; the walls of Caerwent and Silchester, to the third; and the bastioned forts, to the fourth.

The Author's Notes:

1 Roman Fort of Gellygaer, p35.

2 Arch. Aeliana, XXV, p245.

3 Ib. XXIV, plan.

4 Personal observation.

5 Archaeologia, LVII, 340.

6 Personal observation.

7 Soc. Antiquaries Scot. XXXV, p351; XXXV, p167; XXXII, p412; and XXX, p97.

8 Similar stakes have been found at Newstead, and are regarded as tent-pegs.

9 Roman Fort of Gellygaer, p43.

10 See p124.

11 Archaeologia, LII, p752.

12 Ib. LIX, p94; LX, p117, and personal observation.

13 Roman Wall, p316.

14 Archaeologia, LVII, p342.

15 Personal observation.

16 Personal observation.

17 Roman Wall, pp333, 347.

18 Archaeologia, LIX, p87, and LX, p111.

19 Arch. Aeliana, IV (O.S.), p269. Roman Wall, p202.

20 Roman Fort of Gellygaer, p39.

21 Arch. Aeliana, IV (O.S.), p63.

22 Roman Wall, p181. Arch. Aeliana, VII, p171; XXIV, p26.

23 Roman Wall, p317.

24 Soc. Antiquaries Scot. XXX, p101.

25 Ib. XXXVII.

26 Ib. XXXV, p357.

27 Roman Forts on Bar Hill, p22.

28 Ib. XXV, p173.

29 Ib. XXXIII, pp417, 447.

30 Vol. V, p35.

31 Brit. Arch. Assoc., II, p29.

32 Essex Archaeo. Soc. I.

33 E. M. Sympson, Lincoln, p26‑9.

34 Reproduced in Lincolnshire Notes and Queries, VIII, p225.

35 Archaeol. LII, p750; LXI, p474.

36 Ib. LXI, p475.

37 Arch. Cantiana, XXIV. Through a misinterpretation of the remains, it is represented as a double gate, but Mr. John Garstang subsequently corrected the mistake in the Trans. Hist. Soc. of Lanc. and CheshireLII, from the results of the Cardiff excavations.

38 Roman Ribchester, p8.

39 ArchaeologiaLV, p428.

40 Ib. LII, p609.

41 The above statements refer to Britain only. Many town and fortress walls in Italy of the period of the Republic and of Augustus have projecting towers or bastions, and in the east they were of common occurrence in earlier times.

[image ALT: Valid HTML 4.01.]

Page updated: 15 Jul 06