[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 di 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 1
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 3

 p18  Chapter II


Their General Characteristics

The remains we now consider, differ from the last in their stronger construction, but in reality no sharp line can be drawn between the two. A camp intended to serve for winter quarters would be more strongly constructed than one thrown up during a halt in a march; and it is reasonable to think that some of the strongly entrenched posts referred to in the last chapter were designed to last a war of several campaigns. The distinction between forts and fortified towns is perhaps even less marked. It is mainly one of size. Under the former we class the numerous castella designed to hold cohorts or alae, large or small, and the posts or fortlets only large enough for small detachments; and under the latter, the great legionary fortresses, as Chester and Caerleon, and the 'civil' towns, which were more or less planned on the military model. But the two groups are linked together by a few forts of intermediate sizes, perhaps intended to hold double cohorts. Then, again, some of our older towns began as Roman forts, and it is highly probable that the development from fort to town took place in Roman times; it is also probable that some of the 'civil' towns were at first legionary fortresses, the legions having been early removed to points nearer the advancing frontiers. All these military and quasi-military remains, however, have a family likeness, and even if their vestiges are slight and obscure, they can rarely be confounded with the non-military works of the period, or with those of earlier or later times. Our knowledge of the forts and the fortified towns is almost wholly derived from the study of their  p19 visible remains and the evidence of the spade. Their external defences of ditch and rampart have rarely been effaced by the ravages of time. If they happen to be in the heart of a large town, they may be entirely buried under the accumulated débris of successive buildings. Manchester affords an illustration to the point. The position and limits of Mancunium, the fort from which this city took its name, are only indicated by an occasional fragment of its rampart brought to light in some excavation. In more favourable situations, however, the rampart may be a conspicuous object; but the buildings within its line rarely show as more than faint and confused traces.


During the last half century or more, many of these minor strongholds have been systematically excavated. Those which have supplied the most complete plans are Housesteads1 on the Wall in Northumberland, of which portions were explored at different times — by the Rev. John Hodgson between 1822 and 1833, by John Clayton, F.S.A., between 1849 and 1858, and by the Newcastle Society of Antiquaries in 1898, under the direction of Prof. R. C. Bosanquet; Birrens2 in Dumfriesshire in 1895, and Newstead near Melrose, 1905‑8, by the Society of Antiquaries of Scotland, the latter under the direction of Mr. James Curle, F.S.A.; and Gellygaer3 in Glamorgan by the Cardiff Naturalists' Society in 1899, 900, and 1901. Six other forts have been excavated by the Scottish Society with results almost as good — Ardoch4 in Perthshire in 1896‑7; Camelon5 in Stirlingshire in 1900; Lyne6 in Peeblesshire in 1901; and Castlecary7 and Rough Castle8 on the Antonine Wall in 1902‑3. Another Antonine fort at Bar Hill9 was explored by Dr. Macdonald and Mr. A. Park, F.S.A., in 1902. Of the Wall series of forts, Chesters10 and Great Chesters11 have been partially explored, the former  p20 at different times by Mr. Clayton between 1843 and 1890, and some work has been since done there; and the latter, by Mr. J. P. Gibson in 1894. In less degree, Birdoswald12 was excavated by Mr. H. Glasford Potter, F.S.A., in 1850. A small fort at Haltwhistle was excavated by the Newcastle Society in 1907‑8.13 A considerable portion of High Rochester,14 one of the supporting forts of the Wall, was excavated by the fourth Duke of Northumberland, in 1852, and subsequently by the Newcastle Society. This society also laid bare a portion of another fort at South Shields in 1874‑5. A small fort at Hardknott15 in Westmorland was explored by the Cumberland and Westmorland Archaeological Society between 1889 and 1902; in Derbyshire, a fort of similar dimensions, Melandra Castle,16 by the Glossop Antiquarian Society in 1899 and 1900, and several years later by the Manchester Classical Association; and another at Brough by the Derbyshire Archaeological and Natural History Society17 in 1903. In Lancashire, excavations were made on the site of an important fort at Ribchester18 by the Rev. J. Shortt in 1888, and by Mr. J. Garstang in 1898‑9; and the Manchester Classical Society has explored another at Castleshaw.19 At Wilderspool,20 Warrington, the sit of a large fort has been excavated at various times, and especially by Mr. Thomas May since 1895. A small fort at Coelbren21 in Glamorgan was partially excavated by Colonel W. L. Morgan in 1907; and the Liverpool Association for Research in Wales has in hand a larger one at Caersws, Montgomery.22 The Roman remains at Cardiff Castle23 were well revealed during alterations between the years 1890 and 1903; and at the present time the exploration of Pevensey Castle,24 another Roman coast fort, is in progress. Besides these,  p21 a few other Roman forts have been, at one time or another, partially explored.

The Roman forts of this country are of two types, the chief distinguishing feature being the presence or absence of bastions or projecting towers. The larger number are of the non-bastioned type, and there is reason to think that they are the older. The distribution of the streets and buildings within these forts has, in spite of a multitude of differences, a general identity, and it resembles that of the Hyginan camp. In fact, we may regard these forts as translations of that camp into stone or other durable materials, provided we look upon them as free, and not as literal, renderings — a point to be carefully noted. It is evident that while their constructors followed traditional lines, they  p22 exercised discretion as to details. Unfortunately we know practically nothing of the internal planning of the bastioned forts.

Gellygaer supplies an excellent plan of a fort of the non-bastioned type, Fig. 5. It is remarkably simple; and its value is enhanced by the absence of alterations and re-buildings, the remains exhibiting all the signs of being of one design and execution. The general bilateral symmetry — the right and left balancing of parts — will be noted. In this respect, as also in the positions of the four gates and of the streets and buildings of the interior, there is a general resemblance to the camps of Polybius and Hyginus. If the reader pursues the comparison further, he will see in the rounded corners, the narrow intervallum, and the forward position of the via principalis, Hyginan traits, while the approximate squareness of both the fort as a whole and the central block corresponding with the praetorium will appeal to him as a Polybian legacy. The four gates and the many towers at regular distances will recall Josephus' vivid description of a Roman camp. So exactly do his words tally with the remains of this fort, that we might almost suppose him to be describing it in its palmy days — how significant of Roman inflexibility, that the description of a camp at the far east of the empire should so well apply to a fortified post amid the hills of its western fringe!

[image ALT: missingALT]
Fig. 5. — Plan of Roman Fort, Gellygaer. After Rodger.
(100 ft. to 1 in.)

Of the three divisions, the praetentura, the retentura, and the intervening range of the principal buildings representing the praetorium and latera of Hyginus, the first is clearly defined by the via principalis — always a well-marked feature in the forts — which separates it from the other two divisions. But in the absence, which is not unusual, of a thoroughfare behind the principal range, answering to the via quintana, it is not easy to define the limits of the retentura. It is, however, much smaller than the praetentura, and this seems to be a characteristic of early work, for this division is also smaller at Brough, Melandra, and Hardknott; whereas farther north, where the forts are presumably of later date, this is exceptional, the retentura and praetentura being of about equal size at Housesteads, Birdoswald, Birrens, Ardoch, and High Rochester, and the former exceeding the latter at Great Chesters, Newstead, Lyne, and Castlecary.

The position of the via principalis, and consequently of the lateral gates, is unusually backward. The street is a trifle nearer the front of the fort than the back, and this is the position at  p24 Melandra; whereas in all the other excavated forts, with the solitary exception of Brough, it is much nearer the front, the extreme limit in this direction being reached at Housesteads, Great Chesters, and Lyne, while Brough represents the opposite extreme, its via principalis being nearer the back than the front.

The gates, it will be observed, have each two passages and two guard-chambers. The rampart is of earthwork, faced with a strong retaining-wall, while enclosing all is a ditch. The whole plan is slightly askew, a defect due to a faulty setting out of the main lines at the start, and seen in the plans of some other Roman forts.

Our next plan is that of Housesteads, Fig. 6. Its substantial identity with that of Gellygaer will be seen at a glance. There is the same rectangular form with rounded angles, and absence of external projections; the four double gates; and the basements of turrets within the face. The chief streets of the interior also correspond, and divide the buildings into similar groups. The direction of the long, narrow buildings of the praetentura and the retentura is different; instead of being placed cross-wise as at Gellygaer, they here run longitudinally — an unusual feature. The irregular spacing of the partition walls of these and some other blocks bears witness to reconstructions on other than the original lines, but as far as possible these later alterations are eliminated from our plan to avoid confusion. Housesteads, like other Wall forts, but unlike Gellygaer, was long occupied, so long, in fact, that through stress of war, decay and other causes, there was much rebuilding.


[image ALT: missingALT]
Fig. 6. — Plan of Roman Fort at Housesteads. After Dicksee and Bosanquet.
(100 ft. to 1 in.)

The plan of Birrens, Fig. 7, resembles that of Housesteads in its general proportions, but in the arrangement of its internal buildings it is more akin to Gellygaer; the defences, however, contrast with both. Instead of an earth-rampart faced with wall, it is wholly of earth; and instead of a single ditch, there are on the north and best preserved side no less than six. These sweep round the north-east corner, but on the east and south they are now obliterated by the encroachments of the neighbouring streams, and on the west by agriculture. From the analogy of other Scottish forts, it is probable, however, that on these sides the ditches were reduced in number, as the streams referred to formed a natural defence, while on the west there were formerly  p26 indications of a large fortified annexe similar to that at Camelon, which will be described below.


[image ALT: missingALT]
Fig. 7. — Plan of Roman Fort at Birrens. After Barbour.
(100 ft. to 1 in.)

Many of the Scottish forts are remarkable for their massive earth-ramparts and intricate outer defences, especially on the more vulnerable sides; also for their fortified annexes. That on the river Lyne, Fig. 8, is a comparatively simple example. Its situation resembles that of Birrens. It is on a plateau about 100 ft. above the river, which flows at the foot of its west and south declivities, while to the north and east are hollows which were formerly marshes. The fort itself is a short oblong with boldly rounded corners, and is close to the west brow, but is sufficiently set back from the south brow to leave space for a wing-like annexe, which is balanced by another on the north. The rampart is of earth, and external to it is a ditch separated by an interval or berm, both being continuous except at the four gates; but there are additional defences where the 'command' was weakest. Two supplementary ditches sweep round the north-east corner and diverge on the east side so as to leave an intervening wide terrace fortified by an earth-rampart along its exterior margin, while along the opposite edge of the outer ditch is a smaller bank, apparently to increase the height of the counterscarp. Both ditches and banks sweep round the south-east corner as far as the south annexe; the outer ditch is continued to the south-west corner of the fort. Each annexe is defended by a ditch. The internal buildings were arranged as at Gellygaer and Birrens; but while the principal buildings were of stone, those of the praetentura and retentura were of timber.

[image ALT: missingALT]
Fig. 8. — Plan of Roman Fort at Lyne. After Ross.
(200 ft. to 1 in.)

The situation of Camelon is similar to the last. The plateau on which it stands overlooks to the north and east the confluence of two streams, while to the south and west the ground gently slopes away; and the main work is similarly set back from one of these brows — that facing the north — to leave space for an annexe. The fort is almost square and is (or rather was) enclosed by a massive earth-rampart, now almost levelled; but the ditches, of which there were two, an inner narrow, and an outer wide one, ran only along the two sides more distant from the brows, and were continued to the north brow so as to close in the annexe, which was further protected by a rampart. As neither the ditches nor this rampart seem to have been returned  p27 along the edge of the brow, the steepness of the declivity was apparently considered a sufficient defence on the north and east. The annexe was entered by a gate in the above rampart and by the north gate of the fort. At some later date a quadrangular space to the south, and of larger area than the fort itself,  p28 was enclosed by an earth-rampart and ditches. The via principalis of the fort was continued through it, and made its exit by a gate on the south, and it was crossed by an east and west street which passed through a gate on the latter side. The buildings of the main work are remarkable for their arrangement, those of the praetentura, which faced the brow, being longitudinal, as at Housesteads, while those of the retentura were placed transversely as usual. Little is known of the contents of the southern annexe beyond that it contained two large buildings, one of which seems to have been the baths of the garrison.

In the defences of Ardoch we have an extreme example of intricacy. Its west side crests the precipitous banks of the Knaick Water, while to the south and south-east was formerly a stretch of marshy ground. To the north and north-east the ground rises, and in these directions the loss of 'command' is made good by the increased width and intricacy of the artificial defences. The defences on the south are well-nigh obliterated, as they are also to a lesser degree on the west by a modern road. The fort was enclosed by an earth-rampart and ditch, except where interrupted at the four gates. But in addition to these normal defences we have on the east side four additional ditches, which are reduced to three near the south-east corner, and after an irregular interval, an outer bank, which, beginning at the east entrance, sweeps round the north-east and ends at the north-west corner, enclosing in so doing the intricate defences of the north side, Fig. 9. Here we also have five ditches, but they are not all continuous with those of the east side. The innermost follows the rampart as before, but the second, instead of running parallel with it, diverges towards the middle, leaving a terrace-like strip or 'ravelin' of natural surface which widens towards the causeway of the north entrance. Then, at a little distance, leaving another strip which is protected by a small rampart or parapet along its outer edge, are three parallel ditches in close succession, like those of the east side of the fort;25 and finally, the external bank already referred to. The south defences are almost obliterated, but they show less intricacy and width; while those of the west side seem to have been reduced to the rampart and its proper ditch, the river affording a natural defence on that side. The buildings within the fort were almost  p30 wholly of timber, and their remains are very slight and indefinite, but their forms and distribution appear to be normal. Attached to the north end of the fort are the faint remains of a large 'procestrium,' about four times the size of the interior of the main work, which was enclosed with a bank and ditch. This encroaches upon the larger of the two camps described in the last chapter, of which the fort was probably the successor.


[image ALT: missingALT]
Fig. 9. — Plan of Roman Fort at Ardoch, showing intricate defences. Ditches shown in solid black. After Cunningham.
(120 ft. to 1 in.)

The intricate defences of this fort have long puzzled antiquaries. It has been surmised that the Romans simply occupied and modified a British work, but the excavations of 1896‑7 proved that they are wholly Roman. Dr. David Christison admits the difficult of understanding the precise object of some of the details, and attributes them in part to subsequent alterations and additions; but so far from being a "hopeless maze," he sees in the whole a skilfully devised plan. The bewildering complexity of the north-east and north-west corners provided a flanking defence at these angles. The widened ends of the outer ravelin would enable a large number of men stationed there to cross fire with their friends on the extreme ends of the inner line of defence.

Castlecary and Rough Castle on the Antonine Wall, in spite of their disparity of size, have a strongly marked character of their own. Both are applied to the back of that barrier, which thus forms their north fronts, but modified in each case, being of masonry instead of turfwork in the one, and of turfwork of greater width than usual in the other. In both, the rampart of the remaining sides abuts at right angles against the Wall, but the southern corners are rounded, resembling, in these respects, Carrawburgh and the mile-castles on the Wall of Hadrian. The ramparts differ, however, that of Castlecary being of masonry with some evidence of a bank within, while that of Rough Castle is of turfwork like the Wall itself. Each fort has four entrances, the north one being through the Wall; and the military way behind the latter constitutes the via principalis. The outer defence of each is a double ditch, and each has an eastern annexe protected by an earth-rampart and single ditch. The range of principal buildings was of stone, and occupied an unusually large space. The buildings of the praetentura and retentura appear to have been of timber. Castlecary is remarkable for its form, being broader than long, and  p31 still more remarkable are the defensive pits or lilia brought to light during the exploration. They were found just beyond the foot of the glacis of the ditch of the Antonine Wall between the traverse of the north gate and the brink of the valley on the west side of the fort. The pits are 7 ft. long, 3 ft. wide, and 2 ft. 6 ins. deep, and are arranged alternately in ten parallel rows, together forming a band about 200 ft. long and 60 ft. wide.

The Bar Hill fort differs from the last two, and from probably most of the other Antonine Wall forts, in being set back from the Wall. Like both, it has a double ditch on three sides, but on the fourth, that next the Wall, they coalesce into a single one of greater width than either. Its rampart of turfwork is pierced with four gates. Most, if not all the principal buildings were of stone, but those of the praetentura and retentura were of wood. There is no fortified annexe. An interesting feature of the exploration was the discovery of the ditches of a small and earlier fort on the site, with rounded corners and a single gate, and a rampart formed of the upcast from the ditches, of which there were two. The inner ditch closely invested the rampart, but the outer straggled away in a curious manner. On the side opposite the gate was an irregular quadrilateral annexe. This little fort had been long abandoned before the Antonine fort was constructed, and it is attributed to Agricola by Dr. Macdonald.26

The fort at Coelbren reproduces some of the features of the Scottish forts in its earthwork ramparts, double ditch, and traces of an outer glacis. Its exploration did not go far enough to throw light on its internal planning. The remains of one timber building were found, as also traces of gravel, stone-pitched, and clay floors; and it is probable that many if not all of the buildings were of timber.

The excavations at Castleshaw have disclosed the remarkable feature of an inner smaller fortified enclosure, about 200 ft. long and 150 ft. wide, near its south-east side. Both inner and outer forts are of similar shape, and their ramparts are for the most part of piled sods. Within each angle of both ramparts is a patch of roughly laid stones, apparently the foundation of a turret or a ballistarium. Three gates remain of the larger work, but no  p32 opening appears to have been found into the smaller work; but it is probable that the entrance, as well as the fourth gate of the former work, were on the south-east, as the ramparts here are nearly obliterated. Although the two works are of similar construction, it is hardly likely that they were raised together, and the evidence, so far, tends to show that the inner represents a curtailment of the outer to accommodate a smaller garrison. The gates and the internal buildings were of timber, as evidenced by post-holes and the absence of dressed stones.

We need not particularize upon the other forts enumerated on pp19‑20 — Chesters, Great Chesters, Birdoswald, High Rochester, South Shields, Hardknott, Melandra, Brough, Newstead, and Ribchester — beyond remarking that they were all stone forts, like Housesteads and Gellygaer, and of symmetrical form and arrangement; with four gates in their normal positions (those of Chesters and Birdoswald, however, being supplemented with two small gates); with towers set back from the rampart faces; and with one or two ditches of simple type. Wilderspool alone is exceptional in its trapeziform plan and curious internal structures, and it looks like the annexe of a fort rather than a fort itself.

[image ALT: missingALT]
Fig. 10. — Plan of Roman Fort at Haltwhistle Burn. After Simpson.
(80 ft. to 1 in.)

 p33  The small fort on the Stanegate west of Birdoswald, excavated by Messrs. J. P. Gibson and F. G. Simpson in 1907‑8, differed considerably from the foregoing, Fig. 10. It was 212 ft. long, and had a mean width of 185 ft., the east end being wider than the west. The rampart was of earth faced with a wall 3 ft. thick. The west end was set back from the brow of the Haltwhistle Burn, and on the remaining sides was a ditch, which, however, was nowhere parallel with the rampart, the space between the two varying from 18 ft. to 60 ft. On the south and west were two gates of simple construction, each of a single passage of 9 ft. 4 ins. between the jambs, and set back about 8 ft. between two incurvings of the rampart wall. At the west end was a postern, 4 ft. 6 ins. wide, and this and the east gate had been walled up. The buildings were remarkable for their simplicity and arrangement, having little in common with those of the forts described above. On the north side was a long building, 98 ft. 9 ins. by 17 ft. 6 ins., apparently a barrack block. Besides this, there were four small rectangular structures, and a larger one, apparently a yard; and within the north-east corner, a circular oven, 3 ft. 2 ins. in diameter. In the opinion of the excavators, this fortlet was of early construction and of short occupation, and the finds, especially the pottery, confirm this.

Of the bastioned forts, Cardiff furnishes a good example. The early medieval builders of the castle utilized the Roman lines, throwing a great bank over about two-thirds of their circuit, and rebuilding the residue to form the structure known as the 'Ten‑foot Wall.' The discovery of buried Roman masonry was brought about some years ago by the removal of the outer portion of this bank, thus disclosing a corresponding stretch of Roman walling of great thickness and strength, with polygonal bastions at regular distances and the remains of a gate on the north side. From these and less direct indications, it is possible to re-construct the Roman plan, Fig. 11, the actual remains being shown in solid black, and it will be noticed that three of its sides slightly bowed outwards. The gate is of a single span, and presumably there was a corresponding one on the south side, now represented by the medieval entrance to the castle. The central bastion of the east side is externally similar to the rest, but it is hollow instead of solid, and there is reason to believe that it contained a postern. On the west side all the bastions have  p34 disappeared; but the symmetrical planning of the existing remains of the fort renders it probable that this side resembled the eastern.


[image ALT: missingALT]
Fig. 11. — Plans of Bastioned Roman Forts.
(300 ft. to 1 in.)

Richborough in Kent27 (the Rutupiae of the 'Saxon Shore'), Fig. 11, is perhaps the best known of this type of fort, and it was of great importance, as it guarded the chief port of entry into Roman Britain. Its walls are also of great strength, and its remaining gate, on the west side, is of a single span, but not centrally placed. Whether there was a corresponding gate in the opposite or sea-wall, it is impossible to say, as few traces of that wall remain. The bastions appear to have corresponded in number and arrangement with those at Cardiff, but instead of being polygonal, the corner ones are circular, and the lateral, square. The middle bastion of the north side covers a postern; but here again it is impossible to say whether it was balanced by one on the south side. The peculiar feature of this fort is an enormous platform of concrete with a cruciform superstructure, the use of which has not been satisfactorily explained.

Lympne (Portus Lemanis),28 another Kentish fort of the 'Saxon Shore,' has walls still more massive (Fig. 12). It is difficult to make out its exact shape, as these walls have shifted and tumbled about in an extraordinary manner. The south or sea-wall has entirely disappeared; but assuming that it was straight, the fort appears to have been pentagonal, with the two longer sides parallel. The chief entrance was towards the south end of the east side; while at the north salient was a postern, and apparently two others on the west side. The bastions have semicircular fronts and stilted sides.

[image ALT: missingALT]
Fig. 12. — Plans of Bastioned Roman Forts.
(300 ft. to 1 in.)

Burgh Castle29 in Suffolk (the Gariannonum of the 'Saxon Shore'), Fig. 11, also had a sea-wall at the foot of the declivity, of which no remains are visible, but excavations in 1850 revealed the piles on which it stood. The plan, thus completed, is oblong, with the sea-wall longer than the opposite or east wall. The remaining principal entrance is in the middle of the latter wall; and there are indications of posterns in the middle of the north  p36 and south sides. The corners, unlike those of the preceding bastioned examples, are rounded off; and the bastions, which are symmetrically arranged, have bold semicircular fronts with contracted necks, that is, are somewhat pear-shaped.

At Bradwell-juxta‑Mare, Essex (Orthona),30 are some slight remains of another of these coast-forts, Fig. 12, which appears to have been of about the size of Richborough, but was not quite rectangular. The sea-wall, as in the last two examples, has gone. As at Burgh, the corners are rounded off, and there is the opening of a large entrance in the side most distant from the sea. The foundations of two of the bastions, which resembled those of Burgh Castle, remain.

Porchester Castle, near Portsmouth, Fig. 11, — like Cardiff Castle, a Roman fort modified by the medieval castle-builder — is also a coast-fort, but beyond the limits of the 'Saxon Shore.' It is nearly square, with boldly projecting bastions like those of Lympne, which appear to have been symmetrically arranged; and on the side next the sea is a central gate of a single span. Whether this was balanced by another on the opposite side is uncertain, and it is also uncertain whether the remaining sides had posterns.

 p37  Pevensey Castle31 in Sussex (the Anderida of the 'Saxon Shore'), Fig. 11, is another Roman fort which has been utilized as a medieval castle. Its plan presents a striking deviation from the typical Roman form, being somewhat oval in shape in conformity with the configuration of the ground. Its bastions resemble those of Porchester, and its imposing gate — probably the only one — is of a single span and deeply set back between two of the bastions.

At Bittern,32 on a promontory in the estuary of the Irthin, near Southampton, are the remains of a fort as strikingly abnormal as Pevensey. It is triangular, a shape obviously adapted to the apex of the promontory. Little is to be seen of it now; but its strong walls, and the former remains of bastions, warrant its inclusion in the present class.

The remains of similar bastions to those of Porchester, but of less projection, have been found attached to the Roman walls of London.33 The sculptured stones and fragments of architectural details built into them indicate their late rather than early construction.

The reader can hardly have failed to observe that these forts contrast with those of our previous series in other respects than in the presence of bastions. Their walls are remarkable for their thickness and strong construction. Their gates seem never to have exceeded two in number, any additional entrances being small posterns; whereas, in the other series, they were four and exceptionally six. They differed too in their contracted width, consisting, as far as we know, of a single opening. These forts also show a decided tendency to disregard the traditional symmetrical rectangular form. It will also have been noticed that the examples given were estuary or coast defences; this, however, must not be pushed too far, as some inland forts had bastions — the great multangular tower at York was a Roman corner bastion, and the Roman fort at Ancaster has traces of circular bastions capping the corners.

[image ALT: missingALT]

In the first of the two tables next given, the forts are of our first type, and they are selected because their plans are sufficiently perfect to give the particulars required. The length and width  p38 are taken from the outer faces of the ramparts. The position of the via principalis is important, as it carries with it the positions of the lateral gates: the figures express the ratio of its distance from the front of the fort to the total length reckoned as 100. The fact that no remains of turrets have been found in the 'earth' forts does not disprove the former existence of these structures, as they may have been of timber, and so have long since perished. The second table gives a list and particulars of the better preserved bastioned forts.

It will be observed that the areas pass by easy transitions from the 1.2 acres at Rough Castle to the 4.8 at Lyne; while Camelon and Newstead stand apart, the latter after a longer interval than the former. To ascertain how far they may be representative for this country, the writer calculated the areas of nearly seventy forts, including those of the two tables, but excluding the legionary fortresses and the fortified towns. Unfortunately, many of the measurements given are approximate only, and generally it is not stated whether they are within the ramparts or over them, but probably these uncertainties do not materially affect the following results:

Forts     Areas
5 per centum     from 1.3 to  2   acres
79 per centum     from 2.3 to  5.8 acres
6 per centum     from 6.0 to  8.0 acres
10 per centum     from 9.0 to 14.0 acres

Of the second and largest group the preponderating sizes range from 3.3 to 4.5 acres. Of the last it is quite possible that two or three of the largest, ranging from 13 to 14 acres, may be small fortified towns rather than forts, leaving a residue of the same size as Newstead, or slightly smaller or larger. Of the few forts of intermediate size three are of the bastioned type. The inference of it all seems to be that the forts of the second group were designed to hold single cohorts of infantry, normal or extended, or single alae of horsemen; and that those of the last group were designed for double cohorts or alae; while the smallest forts were held by small detachments.  p39 

Fort Dimensions
(English Feet)
Internal Area Position of Via Principalis Rampart Number of Gates Turrets Number of Ditches Retentura compared with Praetentura
'Stone' Forts: Length Width Acres
Gellygaer, Glamorgan 402 385 2.6 46.0 Masonry 4 Angles and sides One Smaller
Brough, Derbyshire 348 287 1.6 55.7 " 4 Angles only " "
Melandra, " 398 368 2.2 48.9 " 4 " " "
Hardknott, Cumberland 375 375 2.4 42.6 " 4 " "? "
Housesteads, Northumberland 609 373 4.4 31.1 " 4 Angles and sides . . . Equal
Chesters, " 573 428 4.5 41.0 " 6 " One Smaller
Great Chesters, " 420 347 2.6 31.6 " 4 Angles only " Larger
Carrawburgh, " 450* 360* 4.0* 33.3* " 4 ? One, two part way "?
Birdoswald, Cumberland . . . 414 4.0 . . . " 6 Angles and sides One "
High Rochester, Northumberland 478 450 3.8 41.0 " 4 Angles only? Three Slightly smaller
Newstead, Roxburghshire 825 720 11.0 35.2 " 4? ? Several Larger
Castlecary, Stirlingshire 365 471 3.6 33.7 " 4 Angles only Two "
'Earth' Forts:
Coelbren, Glamorgan 490 475 3.5 30.9 Earthwork 4 None found Two Smaller?
Castleshaw, Lancashire 399 342 . . . . . . " 4? Foundations? One . . .
Birrens, Dumfriesshire 600 381 3.9 42.5 " 4 None found Several About equal
Camelon, Stirlingshire 602 562 5.8 42.0 " 4 " " Slightly smaller
Lyne, Peeblesshire 580 488 4.8 34.4 " 4 " " Larger
Ardoch, Perthshire 550 490 4.0 39.0 " 4 " " "
Rough Castle, Stirlingshire 272 272 1.2 38.4 " 4 " Two Smaller
Bar Hill, Dumbartonshire 386 399 . . . 40.5 Turfwork 4 " Two, one part way "
(English Feet)
Internal Area
Form Bastions
Cardiff Castle 635 × 603 7.7 Rectangular Polygonal
Richborough Castle 530 × 435 4.8 " Rectangular; circular at corners
Burgh Castle 670 × 420? 5.5? Quadrilateral Pear-shaped
Bradwell-juxta‑Mare 510 × ? ? " Circular or pear-shaped
Porchester Castle 630 × 612 7.5 Rectangular Parallel sides with semicircular fronts
Lympne 790? × 665 9.2 Irregular pentagon "    "
Pevensey Castle 986 × 535? 8.3 Irregular oval "    "

Fortified Towns

Under the settled conditions which followed the period of the conquest, there were three legionary stations — Caerleon (Isca Silurum), the headquarters of the Second 'Augusta'; Chester (Deva), that of the twentieth 'Valeria Victrix'; and York (Eburacum), that of the Sixth 'Victrix' — all conveniently situated as bases for the garrisons of Wales and of the northern frontier. These were fortresses — forts on a large scale, and planned on the same lines, but with larger and more varied accommodation. There were also a number of towns inhabited wholly or mostly by civilians, especially in the south and south-east, and these were also fortified and, so far as we know, were planned on the same lines, some precisely so, others loosely so. To classify the towns as 'military' and 'civil' is convenient, but it is not exact. Some began as military stations, and ended as communities of civilians engaged in the arts of peace. York, long the official capital, had a large civilian population in its fortified suburbs; and it can hardly be questioned that the 'civil' towns included at one time or another — perhaps always, as their fortifications imply — a military element in the shape of a small garrison or guard.

With regard to their external forms, these towns were of two types. The three legionary fortresses, and Colchester, Lincoln, Gloucester, Leicester, Aldborough, Caerwent, Irchester, and some others, resembled the forts, that is, they were oblong. Silchester, Wroxeter, St. Albans, Bath, Chichester, Kenchester, and Corbridge, on the other hand, were irregularly polygonal or  p41 rounded. Several others, as Cirencester, London and Winchester, may be regarded as intermediate, being oblong, but irregularly so; but the latter two may owe their irregularities to extensions in Roman times. Why some 'civil' towns should have been on the military model, may in part be explained by the fact that several — as Colchester and Lincoln — were originally the headquarters of legions; and it is probable that others of the same form had a like origin, although there is as yet no evidence for this. In the earlier stages of the conquest the Roman armies must have been stationed for short periods in other places in the south and south-east than Colchester. Each advance of the conquest carried the scene of military operations farther to the north and west, and it is reasonable to think that the legions would be shifted to new quarters, leaving the maintenance of order in the vacated regions to scattered garrisons. It was only when the conquest was complete that Caerleon, Chester and York became the settled military centres, and remained so to the close of the Roman era. Yet it scarcely accords with facts to make the quadrilateral form a test of military origin. The exploration of Caerwent yielded no evidence of such an origin for that town: on the contrary, all the remains that have been discovered within the walls are those of houses, shops and public buildings appropriate to a 'civil' town.

Caerwent (Venta Silurum) was a small town with a single main street threading it lengthwise and passing through the east and west gates, and a number of minor streets or lanes. Two of these were parallel to the main street, and the rest cut them at right angles, thus dividing the area into rectangular insulae. There were two minor gates, the one on the north and the other on the south, but the one was not opposite the other. Most of the other quadrilateral Roman towns still remain towns, and their present plans are reminiscent of Roman planning. They differ, as a rule, from Caerwent in having two principal streets crossing one another at right angles, as in the forts. This is especially noticeable at Gloucester, where Northgate and Southgate streets represent the chief Roman longitudinal thoroughfare, and Westgate and Eastgate streets the chief transverse thoroughfare. This cruciform arrangement is also well marked at Chester and Lincoln. The relics of the minor streets are well seen on the plans of Winchester and Colchester,  p42 arranged as at Caerwent, and tending to divide the blocks of buildings into rectangular insulae.

The unsymmetrical outlines of the towns of our second type might seem to indicate that these towns were not 'made,' but 'grew,' like most modern towns, and that their fortifications were a late episode in their development. Silchester alone of them has been systematically explored. This ancient town occupied the flat summit of a gentle eminence, and is now covered with fields, with the exception of the parish church and a farmhouse. Its form is roughly octagonal (Fig. 14), but really nine-sided, each side being straight. The fortifications are of earthwork faced with a strong wall, external to which are the remains of two ditches. This line is pierced with four principal gates and several posterns. The streets cross one another at right angles, and the whole plan with its central forum is thoroughly Roman, in spite of the external form.

[image ALT: missingALT]
Fig. 14. — Plan of Silchester (Calleva Atrebatum). (600 ft. to 1 in.)
A, B, C, D, the principal gates; E, the forum; F, public baths; G, hospitium; H, H, H, H, temples; I, church; K, baths; L, amphitheatre.

The exploration, however, has yielded some evidence that when the streets were set-out the site was already occupied. The insulae were not all of the same shape and size, and some of the houses, instead of being built up to the street sides, were set back from them and were canted at various angles. This has been held to indicate that these houses were built before the streets were planned; but the argument loses some force when it is considered that under any circumstances the builders of these houses, not being governed by the streets, were free to choose what aspects suited them best. Still a careful study of the plan shows that it is probable the site was partially built on before the streets were finally set out. The forum and the street facing its entrance are slightly canted to the ESE, and it will be noted that the baths, the temples, and some other buildings are similarly inclined. It would seem that the public buildings and this street were constructed before the general street-system was planned, and that for some reason their exact orientation was disregarded.

So far as can be judged from their remains, the other unsymmetrical Roman towns appear to have resembled Silchester, and like it to have had four principal gates. Roman Chichester was similarly polygonal, and the present streets are reminiscent of a similar rectangular arrangement. Bath was an irregular pentagon.34  p43 Kenchester was an elongated hexagon.35 The excavations at Wroxeter showed the streets around the basilica, and the baths36 were arranged as at Silchester; but the rampart and ditch differ in their irregular curvilinear course. This, in less degree, is also  p44 noticeable at St. Albans, and the streets there are known to have been arranged in two series as at Silchester.

There is both historical and archaeological evidence that many Roman towns in Britain had a British origin. In the vicinity of several may be detected the intrenchments of the British oppida they succeeded. At Silchester most of the British circuit is known, and it enclosed an area nearly three times larger than the Roman. At St. Albans there are traces of a large oppidum, but this, instead of including the Roman town, lies on the opposite side of the river. At Leicester a faint line of entrenchment south of the city may be part of an enclosing work of a similar nature. From historical and numismatic data we know that the first three oppida were respectively the 'caputs' of the British Atrebates, Trinobantes, and Catuvelauni. The British oppidum was a fortified tribal camp and centre, and it probably contained a small settled population whose huts were more or less scattered, but tended to cluster around the house of the tribal chief; but it was not a town as we understand the term.

One of the earliest steps of the Roman conquerors was the establishment of a settled government, and in doing this they appear, as a rule, to have adopted the British tribal territories as their units of administration, and with the territories, the tribal capitals. Thus far they kept up a link with the past, and to this was probably due, in great measure, the rapid acquiescence of the natives in the new conditions. How far they modified the old machinery is uncertain; but undoubtedly they gave it a Roman form. With the adoption of the capitals would follow the remodelling of them on Roman lines, and the erection of offices for carrying on the new administration. Silchester affords a good example of the result. Although a certain amount of trade was carried on within it, it was not strictly a commercial centre. It was not full of shops like Pompeii. The houses were of a goodly sort, with intervening yards and gardens. Its population probably never reached 3000, and its central forum and administrative buildings were altogether on too large a scale for the municipal needs of so small a town. It appeals to one as essentially a residential town, and perhaps many of the residents were the officials of the civitas or canton. The roads radiating  p45 from it afforded easy and rapid communications with all parts of the territory it controlled.37

The towns were not all of the same constitution and rank. St. Albans was a municipium, and Colchester, Lincoln, Gloucester, and York were coloniae; and possibly some other towns shared in the high privileges of these. But however important these towns were at first, they were gradually overshadowed by others which had not these privileges, especially by London, then, as now, the commercial metropolis, until at length the decree of Caracalla, in extending the privilege of Roman citizenship to all the free inhabitants of the Empire, constituted every provincial town a municipium in fact, if not in name. Whatever advantage the old municipia and coloniae now retained, would probably be one of rank and honour only, the distinction between them and the others being of a similar nature to that between our 'city' and 'borough.'

To return to the fortifications of the towns — it can scarcely be questioned that those of the quadrilateral towns, like those of the forts, were raised, or at least provided for, at the outset; and this is presumptive evidence for the contemporaneity of those of the unsymmetrical towns. It is true that the Icenian revolt found Camelodunum unfortified, but the statement of Tacitus implies that this was exceptional, the result of neglect. Silchester offers something towards a solution. Its circumvallation approximates to the configuration of the site. The road from the west bends to due east upon entering the town through the west gate, showing that this gate marks the limit of the town on that side, when the streets were set out. These, of course, do not amount to conclusive proof, but they favour the view that the circumvallation and the planning of the interior were simultaneous. By adopting the polygonal form, the engineers were able to enclose a maximum of space, and to obtain a maximum of 'command' for their rampart. Had they adopted the rectangular form, the enclosed space would have been smaller, unless a portion of the slope of the hill had been included; but it would have left portions of the plateau unenclosed. The irregular curvilinear form at Wroxeter has not a Roman look, and in this  p46 case the Romans may have simply utilized the older lines of a British oppidum which happened to be of suitable size for their purpose.

It thus appears that the towns were fortified early, and not late; but there is good evidence that long after their foundation, their defences were modified and strengthened. The excavations at Caerwent have shown that the rampart was originally of earthwork only, and was afterwards faced with a massive wall, and it is almost equally certain that this was also the case at Silchester. That the existing Roman wall at Chester was also a subsequent work is proved by the large number of Roman tombstones and other worked stones used in its construction; and these show that it could not have been erected before the middle of the second century. Carved and sepulchral stones were similarly used in the bastions of the Roman walls of London. The great thickness and other peculiarities of the town walls also indicate their late, rather than early, date, as will be more fully discussed in the next chapter. This strengthening of the fortifications of the towns is generally assigned to the close of the third and early part of the following century. The motive could hardly have been the fear of foreign invasion. It must have been a growing sense of internal insecurity, such as the lawlessness and rival factions of the era of pretenders to the purple between the death of Severus and the strong rule of Diocletian would give rise to.

The Author's Notes:

1 Roman Wall, Bruce, Arch. Aeliana, N.S. XXV, 193.

2 Proc. Soc. Ant. Scot. XXX, 81.

3 Roman Fort of Gellygaer, J. Ward.

4 Proc. Soc. Ant. Scot. XXXII, 399.

5 Ib. xxxv, 329.

6 Ib. xxxv, 154.

7 Ib. xxxvii, 268.

8 Ib. xxxix.

9 The Roman Forts on the Bar Hill.

10 Arch. AelianaIII (O.S.), p142; VII, p211; XIII, p374; XXIII, p268.

11 Ib. xxiv, p19.

12 Arch. Aeliana, IV (O.S.), p63.

13 The Roman Fort at Haltwhistle, Gibson and Simpson.

14 Arch. Aeliana, N.S. I.69; Bruce, Roman Wall, 315.

15 Trans. Cumb. and Westmor. Arch. Soc. xii, 375.

16 Melandra Castle, ed. by R. S. Conway.

17 Jour. Derbysh. Arch. and Nat. Hist. Soc. xxvi, 177.

18 History of Ribchester, Smith and Shortt. Roman Ribchester, J. Garstang.

19 First Interim Report, 1908.

20 Warrington's Roman Remains, T. May.

21 Arch. Camb., 1907, p129.

22 Report not published.

23 Archaeologia, LVII, p336; Arch. Camb., 1908, p29. J. Ward.

24 Arch. Jour. lxv, p125.

25 These were originally continuous across the entrance.

26 A similar early fort has been found on the site of the great fort at Newstead, also attributed to Agricola.

27 Roach Smith, Richborough, Reculver, and Lymne. Fox, Archaeological Journal, 1896.

28 Roach Smith, Excavations on the Site of the Roman Castrum at Lymne, 1850.

29 Personal Observation, and Remarks on the Gariannonum of the Romans, J. Ives.

30 Coll. Antiq. vii, 155. Arch. Jour. xxii, 64; xxiii, 60.

31 Munimenta Antiqua, E. King, ii, p38.

32 Englefield, Walk through Southampton, 2nd ed., 81. Proc. Soc. Ant. 2, xix, 56.

33 On a Bastion of London Wall, J. Edward Price.

34 Scarth, Aquae Solis.

35 Arch. Journal, xxxiv, p354. Vict. History, Herefordshire, i. 175.

36 Wright, Uriconium.

37 A monument was found at Caerwent in 1902, with an inscription to the effect that its erection was decreed by the ordo or senate of the civitas of the Silures, of which Venta Silurum was the chief town or capital.

[image ALT: Valid HTML 4.01.]

Page updated: 3 Mar 18