[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 3
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 5

 p80  Chapter IV


Their Internal Buildings and 'Suburbs'

Our knowledge of the internal buildings of the castella is confined to those of the earlier or Hyginan class, as the exploration of the bastioned class has rarely gone beyond their defences. Sufficient has already been said about their general planning in Chapter II to allow us to describe the different buildings or groups of buildings without further introduction. Comparison of the more complete plans of these forts proves that with comparatively few exceptions the buildings are resolvable into four types, of which the first three in the following summary nearly always occupied the middle zone of the fort, and the last, the remaining spaces on either side:—

(1) A rectangular building of remarkably constant plan on the side of the via principalis next the back of the fort. Its wide entrance faced the front gate, and opened into a porticoed courtyard. Behind this was a narrow space extending the full width of the building, and usually regarded as a second yard, and at the back of all was a row of offices, of which the middle one was the most important.

(2) Two or more strongly constructed oblong buildings, almost invariably with buttresses. They do not appear to have been divided into rooms, and their floors rested upon dwarf walls or pillars, the spaces between which communicated with openings in the side walls to allow of the circulation of air through them in order to keep the floors dry. These buildings occurred singly and in pairs, and were usually near the lateral gates of the fort.

(3) A house-like building — in several instances two — divided into a number of rooms, which sometimes surrounded a courtyard.

 p81  (4) A number of long, narrow buildings, usually divided into rooms, and symmetrically arranged across the praetentura and retentura; but sometimes they were arranged longitudinally instead.

Little can be gleaned from ancient writers and inscriptions as to the uses of these different buildings; but a consideration of their plans, their distribution compared with the tents of the Hyginan camp, and the varied needs of a garrison establishment, provide a reasonably satisfactory solution. The central building was certainly the headquarters; and almost equally the buttressed buildings were granaries. The house-like structures can scarcely be regarded otherwise than as the residences of the commandants and probably the chief officers of their staffs. Accommodation for the soldiers and their petty officers must have been provided, and as the long buildings both in form and distribution answer to the lines of tents in the Hyginan camp, there is little doubt that many of them were barracks.

These, however, neither exhaust the buildings of a fort nor the needs of a garrison. On all the more complete plans may be noticed other structures that cannot be classed with the foregoing. On the other hand, each fort was the scene of many necessary operations — the cornº had to be ground and the daily food prepared, and there must have been repairing shops of various kinds, as smithies, armouries, joineries, and so forth. In cavalry forts, and those containing both infantry and cavalry, the stables must have been an important element; and perhaps in most of the infantry forts a few horses were kept for scouting purposes and dispatches — and horses imply the storage of fodder. Among the minor structures would be latrines, cisterns for the storage of water, ovens and other cooking arrangements, wells, drains, etc. Small baths have been found in some of the forts, but those for the use of the garrisons were almost invariably outside the walls.


This building is the 'forum' of some antiquaries, and the 'praetorium' of others. Each term has a certain appropriateness.  p82 It is, on the one hand, forum-like in plan: on the other, it occupies the position of the praetorium in the camps of the classical writers, but it differs in not being the residence of the commander, nor, in fact, a residence at all. This, however, is not of much moment, as the praetorial space in the camps contained the tribunal, the altars, and the auguratorium, which would constitute it the headquarters. There is no evidence that the central building of the forts was anciently known as the 'praetorium.' There is fair evidence, however, that it was called the 'principia.' Three fragments of an inscribed tablet, found within the entrance of the central building at Rough Castle, recorded the erection of the principia by the Sixth Cohort of Nervians in the reign of Pius. Two other inscriptions have been found, the one at Lanchester1 and the other at Bath, recording the restoration of ruined principia.

The headquarters at Chesters (Fig. 28) was one of the largest in Britain, and architecturally one of the finest; and its remains have the advantage of being open to inspection. The courtyard was paved and had a marginal stone gutter with an outlet into a drain in the upper left-hand corner;º while in the opposite corner was a well. There were three porticoes, of which the lateral had square piers, while the passage-like front one was separated from the yard by a wall pierced with a large central opening and two smaller ones, all probably arched. Along the back of the courtyard stretched the front wall of the second division. It had five openings, all probably also arched, of which the end ones were smaller than the others and provided direct access from the porticoes. The space behind extended the full width of the building, and had a portico or aisle supported by four oblong piers, at each end of which was an external door. This space was paved like the yard, but, unlike it, had no marginal gutter. Along the back were five rooms. The opening of the middle room was certainly arched, and within were the steps into a vaulted chamber under the adjacent room on the left; but neither the vault nor its access is part of the original structure. The openings of the adjacent rooms were probably also arched, while the end rooms were entered from these. That to the right of the middle room had a central square of flagged paving.

[image ALT: missingALT]
Fig. 28. — Plans of Principia or Headquarters.
(50 ft. to 1 in.)

 p83  The corresponding building at Birrens2 (Fig. 28) was smaller. The external walls were buttressed. The yard was paved and contained a well. On each side was a wall of secondary work which probably replaced a row of piers, thus converting the lateral  p84 porticoes into closed rooms. There are no signs of a front portico, and this probably explains the absence of a gutter on this side. But near the back of the yard were found four base-stones to receive the posts of a verandah, and as the gutter passed along the foot of the arcade between the two main divisions, it is almost certain that this verandah was a late introduction. The arcade was of seven arches, supported on square piers with moulded caps. The space behind was undivided, had external doors, and was roughly paved, but without marginal gutters. The five rooms resembled those of Chesters, except that the end one on the left was entered directly from the space in front. The middle room contained a sunk chamber lined with large slabs and entered by steps, and broken window-glass was found in it. The room to the left had a square shallow sinking in its floor, and the surrounding pavement was much worn; while each of the end rooms had a square void space in the paved floor.

The headquarters of Gellygaer3 (Fig. 28) resembled the last, but was of simpler construction and not cumbered with late work. The yard was gravelled, with a well on the left, and on the opposite side an enigmatical pit, which appears to have been refilled without delay. On all four sides were the square foundations of the roof-supports. There were three porticoes of equal width, and they, as also the cross-space, had gravel floors kerbed off from the yard. The rain-water escaped by a drain in the upper left-hand corner of the latter. The five rooms at the back exactly resembled those at Birrens, but the middle one projected beyond the general line of the back of the building. The floors were of beaten earth, except that of the middle room, which may have been of timber. The end room on the left had a small sink which emptied into a drain passing through the side wall, and the two rooms at the right had central patches of flagstones, both discoloured by fire. The various roofs had been covered with red tiles; and it is worthy of notice that the outside wall of the first division is thinner than that of the second.

The remains of the headquarters at Housesteads4 (Fig. 28), which are open to view, prove that the building underwent many alterations, and that the porticoes had been walled up as at Birrens. The earlier work has been so much interfered with,  p85 that the recovery of the original plan is difficult. Two plans are given, the second being that of Prof. Bosanquet, who excavated the site in 1898, and the first, that of the writer, who examined the remains several years later. The former resembles that of the Chesters headquarters, except that the second division is separated from the first by a wall with only a central opening. The latter shows this wall as the front of a portico, balancing that of the opposite portico, and the courtyard is shown as surrounded by a continuous portico of equal width throughout. There is no doubt that in the latter period of the building, there was a wall as indicated by Prof. Bosanquet, but to the writer it seemed to have embedded in it the bases of piers corresponding with those of the opposite portico. The courtyard was paved, but the raised pavement of the ambulatory encroached upon it about 2 ft. 6 ins., thus carrying the stone gutter about 3 ft. in advance of the colonnades and showing that the portico roofs overhanged to that extent. This gutter was returned along the farther side of the courtyard at the same distance from Prof. Bosanquet's wall, showing that along that side there was also an overhanging roof.

Passing into the second division, the cross-space, which had a roughly patched pavement without a marginal gutter, and a door at each end, was, according to the one plan, undivided as at Gellygaer, and according to the other it had a portico or aisle as at Chesters. The five rooms at the back contain much late work, and the openings have been narrowed or entirely blocked; but it is easy to disentangle their original planning, and this closely resembles that of Birrens. The middle room is especially interesting. Its opening is about 12 ft. 6 ins. wide, and it was originally arched. The sill is of two long stones, and its upper surface, except for 4 ft. 7 ins. in the middle (where it forms a much worn step) has a shallow chase, 8 ins. wide, with a plinth moulding on the outer side, and this chase is continued up the face of the remaining jamb. These details are well seen in Fig. 30. There is no doubt that these chases held a wall or screen of stone or timber with a central opening of the width of the step. From the worn condition of the plinth it may be inferred that this structure was either low or of open-work, and that the wear was caused by the feet of those who stood in front and looked into the room. In Prof. Bosanquet's restoration  p86 it is shown as a parapet with a central opening, which probably was closed with a gate. The floor of the room was originally of opus signinum, which was afterwards covered with broken building stone overspread with clay. Those of the other rooms were of clay upon stone chippings, and the clay was more than once renewed. None of these rooms contained a vault.

[image ALT: missingALT]

[image ALT: missingALT]

Of the headquarters at Hardknott5 little else than foundations remain. The plan shows the sleeper-walls of three porticoes; an unusually narrow cross-space; and instead of five rooms behind, three, but the end ones are unusually long and may have been subdivided into two, each by thin partitions that have disappeared. At Melandra,6 there were also three rooms, and the positions of the doors tend to confirm a similar subdivision of the end rooms. The corresponding building at High Rochester appears to have resembled that of Chesters, except that its cross-space lacked the aisle. The rooms at the back appear to have been five as usual, but only the two to the left were clearly defined. Of these the end one had a door from the cross-space, and the next a hypocaust. The most interesting feature was the vault in the middle room. It was of strong construction, with a flagged floor, and was reached by a flight of steps, at the foot of which the chamber was closed by a large stone slab which moved in a groove upon iron wheels. In the corner of the chamber at one side of the steps was an arched passage, which unfortunately was not explored. The remains of the headquarters at South Shields were also imperfect and complicated with late alterations. The courtyard had three porticoes with a gutter in advance of their fronts as at Housesteads, and, as in Prof. Bosanquet's plan of these headquarters, a wall with a central doorway seems to have separated the cross-space, which had no side doors. The middle room had a vault entered by steps as at High Rochester.

The headquarters at Newstead, to judge from the small plan published in an interim report, resembled that of Chesters, but was larger. "Alterations which had evidently been made on the building, suggested that it had been used during, at least, two distinct periods. In the earlier of these, the building followed the normal type. Entering from the via principalis was the outer courtyard, surrounded by an ambulatory or cloistered  p87 walk. Farther from the entrance was the inner courtyard (our 'cross-space'), on the west of which were situated the usual five chambers." Among the latter alterations were the introduction of a vault or strong room beneath the middle room, and the construction of a great hall, about 154 by 50 ft., over the via principalis in front of the building, to serve probably as a drill hall. The chief feature of the scanty remains of the headquarters at Great Chesters7 was the arched vault which closely resembled that of Chesters. At Brough,8 the remains were even vaguer, but on or about the site of the middle room at the abstract was found a 'pit' of well-constructed masonry, 8 ft. long, 7 ft. wide at one end, and 5 ft. at the other, and 8 ft. deep, with a flight of stone steps on one side to the concrete floor. The steps, however, were not part of the original construction, and when they were inserted, a small square pit was cut through the floor apparently to collect any water that drained into the chamber, as the remains of a wooden bucket were found in it.

The headquarters at Rough Castle9 were remarkable for their narrow form and unusual divisions, due to the narrowness of the fort. The plan is too imperfect to make out the details with certainty. The courtyard was small and was paved, and appears to have been flanked by two relatively wide porticoes. There were three rooms at the back of the building, and the middle one contained a small pit lined with flagstones, 4 ft. by 2 ft. 3 ins., and 2 ft. 6 ins. deep. The intervening space was apparently subdivided by a cross-wall with a central opening. The corresponding building at Camelon10 contrasted with this in its great width, but its remains were very vague. The courtyard had two lateral porticoes of unusually great width, and the cross-space was long and narrow.

The reader will have observed that while the headquarters' buildings agreed in their general planning, they differed in their general proportions and in minor respects. The normal arrangement seems to have been a courtyard with three porticoes, and a second space with five rooms along the back. Several, however, had only lateral porticoes, and Housesteads possibly had porticoes on all four sides of its courtyard. At Hardknott and Melandra there were three rooms at the back, but the end rooms were long  p88 enough to have been subdivided; this, however, could not have been the case at Rough Castle. The space between the courtyard and these rooms — generally regarded as a second courtyard — was invariably narrow, varying from 13 ft. at Hardknott to 26 ft. at Chesters. At Gellygaer, Rough Castle, and probably at Melandra, Hardknott, Camelon, and Newstead, it was entered only from the courtyard and its porticoes; while at Chesters, Housesteads, Birrens, and High Rochester, lateral doors provided additional means of access. At Chesters, Housesteads (according to Prof. Bosanquet), and probably Rough Castle and Newstead, it had an aisle or portico next the courtyard.

The rooms at the back of these buildings were undoubtedly administrative. The middle one was the most important, and Housesteads supplies a hint how its open front was treated. That it was the sacellum — the place where "honours were paid to the standards which were exhibited within it, to the Genius of the regiment, and to the Imperial house11 . . . the discipline and esprit de corps of the Roman army being closely bound up with the worship of the standards, and the worship of the emperor" — scarcely admits of a doubt.

It was also the treasury. We have observed that in many of our examples an underground cell or vault was provided, and that it was of late work. From this it would seem that in the declining days of the empire the growing lawlessness necessitated a stronger protection for the treasure (in earlier times probably kept in a strong chest) than that afforded by the sanctity of the spot. It is impossible to say what the remaining rooms were used for, but the recurrence of the same number in most of the forts implies that each had a customary and definite administrative purpose. Those next the sacellum had, like it, wide openings, but how these were treated is uncertain. The narrow doorways of the end rooms suggest some degree of privacy, and this is emphasized in those cases where the doorways opened into the contiguous rooms.

With regard to the general construction of these buildings, the first division involves no difficulty. The yard was open to  p89 the sky, and the portico roofs sloped towards it, as is sufficiently proved by the marginal gutters. That the five offices at the back were roofed can hardly be questioned, but whether by a single longitudinal roof or several transverse ones, is uncertain. The broken window-glass in the sacella at Gellygaer and Birrens points to windows or skylights.

The 'cross-space,' as we have so far termed it — the 'Querhof' of the German antiquaries — is generally regarded as an inner yard, but there is reason to think that it was roofed, at least in our cold country, in which case 'cross-hall' would be an appropriate designation. Its narrow width is suggestive that it was roofed; and it will be recalled that while the front yard in our better preserved examples of these buildings was well paved or gravelled and had marginal gutters to carry away the rain-water, this space was less solidly floored and was not provided with gutters. It has been noticed in several instances, that the gutter was returned along the back of the yard, and this suggests the eave of a roof above. According to the current view, this roof at Chesters could only have been that of the portico of the inner yard. But a portico roof sloping away from its open side is contrary to experience, as the motive of such a structure is shelter, and a lofty front would unduly expose the covered walk to sun, rain, and wind. In Prof. Bosanquet's reconstruction of this portico at Housesteads the roof slopes to the inner court, thus rendering the gutter behind meaningless. The evidence of Gellygaer is still more emphatic. The row of piers or pillars along the back of the yard must have supported a roof, but as there was no inner line of piers as of a portico, the opposite edge of this roof must have rested on the front wall of the five rooms behind. Further, the greater thickness of the external wall of theº second division of the building tends to confirm this, as the roof of a cross-hall would be both heavier and loftier than those of the yard porticoes, and if the builders made any difference in the thickness of the external walls, they would naturally apportion it to the weight to be sustained. Birrens is equally to the point. The cross-space, like that of Gellygaer, had no inner portico or aisle; yet the gutter along the foot of its arcade implies a roof, and a roof in this position could only have spanned the intervening space.

These headquarters' buildings bear too close a resemblance to  p91 a well-known type of forum group, of which those at Silchester and Caerwent are good examples, to be a mere coincidence. The cross-halls answer to their basilicas, and the rooms behind to their curiae, where justice was administered and public business transacted.12

Fort Dimensions
(English Feet)
Yard Porticoes Cross-Hall Administrative Rooms
Length Width Aisle External doors Number 'Strong-room'
Chesters 125 86 3. Front narrow Yes Yes 5 Yes
Birrens 78 68 2. Lateral only No " " "
High Rochester 78 73 3. Front narrow "? "? " "
Housesteads 90 76 3? 4? Equal width ? " " No
South Shields 90?? 79? 3. Front slightly narrower No? No Space for 5 Yes
Gellygaer 80 78 3. Equal width " " 5 No
Hardknott 70 70 3. Front slightly narrower " " 3. Space for 5 "
Melandra 73 72 Not traced "? "? " "
Great Chesters ? 78 " "? "? 5? Yes
Camelon 92 120 2. Lateral only. Wide? " ? Space for 5 No
Brough 60? 85 Uncertain. Lateral only? "? Yes? " Yes
Rough Castle 75 43 2. Lateral only Yes? No 3 "
Newstead 132 103 3? 4? Equal width "? "? 5 "
Lyne 95 104 Uncertain No "? Space for 5 No
Bar Hill 83 77 2? Probably lateral only " " 3, possibly 5 "
Castlecary 99? 85 Uncertain ? "? 2?? "

Granaries or Storehouses

[image ALT: missingALT]
Fig. 31. — Plans of Horrea or Granaries.
(50 ft. to 1 in.)
These were the strongest of the internal buildings, and their remains are usually the least obliterated (Fig. 31). At Gellygaer no other buildings were more encumbered with fallen débris, or showed more distinctly on the surface, indicating that they were not only strong, but lofty. In form they were oblong, and varied more in length than in width, the former ranging from 54 ft. at Hardknott to 130 ft. at Newstead, while the latter seldom exceeded the limits of 22 ft. and 25 ft. Their walls were rarely less than 3 ft. thick. They were single or double. Gellygaer and Newstead had two single blocks each, Birrens three, and on the plans of Camelon, Lyne, Rough Castle, Castlecary, one is shown in each; but as these plans are incomplete there may have been more. The same applies to Chesters and Corbridge,13 where two stood side by side, but wholly unconnected and somewhat different, indicating probably  p92 a difference of age. High Rochester possessed two double blocks; and Housesteads and Hardknott one each. At Birrens the largest block had a transverse wall across the middle, and so may be regarded as a pair placed end to end.

Two other peculiarities distinguish these buildings — their provision for raised floors and lateral openings between the buttresses. The actual floors have disappeared except at Corbridge, but their supports remain to a greater or less degree. These were of two kinds, parallel dwarf walls and pillars. The former were the more usual. At Housesteads, Chesters, Newstead, Birrens, Bar Hill, Rough Castle, and Corbridge, they ran lengthways about their own width apart, thus dividing the space into a series of narrow channels, and in the last three they were interrupted at intervals to form a series of narrower cross-channels. At Hardknott, each had a single longitudinal dwarf wall, midway between the external walls. At Gellygaer, these walls were transverse, 6 ft. apart in the one building, and 9 ft. in the other, and all had central openings 3 ft. wide. At Lyne, they were also transverse, and apparently widely spaced, but too few remain to determine the width of the intervals. The pillared type occurs at Housesteads, South Shields, and Castlecary, the pillars of the first two being squared blocks of stone, and in the last, rough boulders.

The openings only remain where the external walls are sufficiently high to show them, and they appear to have risen from the ground-level. At Rough Castle, they were slits, 5 ins. wide externally and 9 ins. high; at Castlecary, 6 ins. wide, and remaining to a height of 3 ft.; and at Corbridge, nearly 2 ft. high and about 1 ft. wide, with a central stone mullion, the ends of which were let into the sill and the lintel. At Gellygaer, the openings were not splayed as in the foregoing, and were originally 3 ft. wide, with sills on the street-level, but were subsequently reduced to 2 ft. by the insertion of masonry cheeks. The original height is unknown, but this must have varied, as, owing to the general slope of the ground, the streets were on different levels. At Birdoswald, the buttressed wall of a granary at least 92 ft. long was discovered in 1859 near the south side of the fort, with narrow slits or loopholes between the buttresses.

[image ALT: missingALT]

Birdoswald: the two horrea, or granaries, now fully excavated. We see one in full; to the right, a glimpse of the other.

Photo © Jona Lendering 2009, by kind permission.

At Corbridge alone, portions of the raised floor remained.  p93 The flooring flags were large enough to span the intervals of the dwarf-walls, and were on the level of the lintels of the side openings, the channels between these walls being 2 ft. 7 ins. high in the one building and rather less than 2 ft. in the other. In all the other granaries, with the exception of those of Gellygaer and Hardknott, the intervals between the supports could similarly be spanned with flagstones, but the wider intervals at these two places would require timber beams to carry the floors. The basements at Gellygaer could hardly have been less than 3 ft. high. These hollow understructures were not hypocausts, as in no instance have stoke-holes or wall-flues been found in connexion with them; in fact, the lateral openings would render them useless as hypocausts. The presence of sooty earth at Camelon, and of charred wheat at Birrens, may indicate nothing more than that the buildings were destroyed by fire. The only feasible explanation is that their function was to keep the floors above dry by the free circulation of air through their spaces.14

The remains of the doors have been found in several of these buildings. Each component of the pair at Housesteads had, at the end farthest from the via principalis, a wide door reached by steps, of which the pivot-sockets and bolt-holes remain. The component containing the kiln (p104) had in addition a door next the street, but this was a late insertion, probably contemporary with the kiln. The doors at Gellygaer were obliterated, but their positions were indicated by the remains of porches, of which each building had one at each end. These structures appear to have been of timber, each erected upon a low platform of the width of the building, and edged with a built kerb covered with flagstones. The kerb of the longer side or front supported four posts, and the spaces between these were fenced or otherwise filled in, for the lateral kerbs alone were worn, showing that the porch was entered on those sides. Within each was found a loading platform built against the main fabric. At South Shields, also, no actual traces of doorways were found; but a row of pier foundations at one end of the double block may have related to a porch. At High Rochester and Castlecary the doors remained, and they opened upon the via principalis; and this was  p94 the position of the doors at Rough Castle and Lyne. At Chesters, Birrens, Hardknott, Camelon, and apparently Newstead, there were neither remains of doors nor indications as to their positions.

By thus piecing fact with fact we have reconstructed these buildings to the levels of their raised floors; we can do little more than conjecture what their superstructures were like. Their basements supply no evidence for divisional walls above them. It is probable that each building — or if more than one storey, each storey — was a single spacious apartment. One reason for the buttresses is well seen at Gellygaer. As the original openings were wider than the intervening masonry, they would represent a line of weakness at the base. To ensure stability, the builders did what a modern architect would do — they strengthened the intervening wallings by buttresses. Whether these were continued up to the roof, pilaster-wise, or were sloped off at a lower level like a gothic buttress, is uncertain. But if the superstructure was of more than one storey, the lower storey would certainly have lateral openings if only for ventilation; and if of a single storey it is likely that this end would be attained in the same manner, rather than by openings in the roof. These openings would certainly be placed vertically over those of the basement, and in this case it would be constructionally necessary that the buttresses should be carried to the full height of the walls, and bear the weight of the main timbers of the roof. At Gellygaer, the roofs were covered with red tiles, and so would be of great weight. Probably they overhung to the full projection of the buttresses, and thus materially helped to keep the walls dry and to ward the rain off the openings. Whether these buildings were of more than one storey is not easily answered. If there were upper floors, these in the wider buildings would surely have required supporting pillars or posts; but the remains show as little evidence for them as for divisional walls, and no trace of external or internal stairs, by which such floors would be reached, has been found. In the absence of evidence to the contrary, it is safer to conclude that these buildings were of one storey, lofty, and open to their roofs.15

In the following table the first column gives the number of  p96 granaries that have been found in each fort, and this in most cases certainly represents the actual number that the fort contained. The least certain in this respect is South Shields, which probably contained two pairs, as remains of a buttressed wall have been found on the other side of the head-quarters. The dimensions in the second column are external, but not including the buttresses. (p95)

(English Feet)
Lateral Openings
Transverse Walls
Longitudinal Walls
Not found
South Shields
Pillars in one; Longitudinal Walls in other
Side by side
Longitudinal Walls
High Rochester
2 Pairs
71 & 75
End to end
Longitudinal Walls
"    "
Not found
"    "
"    "
Transverse Walls
Rough Castle
Longitudinal Walls
Bar Hill
Longitudinal Walls in one; no supports in the other (?)
Not found
Side by side
Longitudinal Walls

Official Residences

On every plan of a fort that shows the buildings about the head-quarters, there is one which vies with that structure in size, and sometimes is even larger. The planning is less stereotyped than those of the buildings already considered; but the distinguishing feature is ever the division into a number of rooms, which gives it a house-like character, and for this reason it is regarded as the commandant's residence; but it may also have included rooms for the chief members of his staff. On at least two of the plans, those of Housesteads and Birrens, two of these buildings can be distinguished — a smaller on the former behind the headquarters, and a larger on the latter on the opposite side of the via principalis.

The plan of the Gellygaer example (Fig. 32)16 is easily made out, although the remains are scanty. The house was entered from the street, not directly, but through a corridor or portico extending along the front and open at the ends; and the door within this opened into a short passage which led to the corridor of a small courtyard, around which were the rooms. These were mostly long and narrow, and some of them may have been divided by partitions which have disappeared. Little further can be gleaned, except that the upper work was probably of timber, as very little stone débris was found on the site, while the sprinkling of red roofing-tiles and broken window-glass points to tiled roofs and glazed windows. The plan is that of a house of the so‑called 'courtyard-type.' The corresponding building at Lyne17 seems to have been similar, so far as can be judged from its scanty remains. The two Housesteads' examples18 were also of the same type, and in each the rooms were numerous.  p97 In the smaller and better preserved building (Fig. 32) the corridor passed along three sides of the courtyard, and had a short passage from the street, as at Gellygaer.

[image ALT: missingALT]
Fig. 32. — Plans of Commandant's Houses at Gellygaer and Housesteads
(30 ft. to 1 in.)

The smaller house at Birrens19 was of different type. The plan shows a long passage intersecting it, the one end being the street entrance. This divides the building into two unequal  p98 portions, which may have served different purposes. The larger block on the opposite side of the via principalis apparently had a small central courtyard, and the western group of rooms seems to have formed an independent set from the rest, but communicating with the presumed courtyard by an internal passage and small door. This block is balanced by the large double storehouse on the other side of the longitudinal street, and as the garrison included a detachment of horsemen, it is possible that both, in common with the other buildings of the praetentura, related to that detachment.

At Chesters, a number of rooms20 between the headquarters and the east rampart have been opened out, but they extended farther to the north and south. So far as the plan goes, they appear to have belonged to two if not three blocks, of which that next the rampart has all the appearance of being a bath-house. In the same position at Great Chesters, a large house-like building21 surrounded a paved courtyard, in which were found some socket-stones for posts, apparently the supports of a portico. Several of the rooms were heated by hypocausts, and a wing next the headquarters seems to have been baths. At the extreme left of the headquarters' range at Camelon22 was a large building of the class we are considering, which, to judge from the small plan in the report, had a number of rooms arranged around three sides of the upper half of a courtyard, and in the lower half an undivided transverse apartment or space with an apse towards the rampart, and next the via principalis what may have been a large yard.

At Rough Castle, the whole space between the granary and the west rampart was occupied by a large rectangular structure,23 but its remains were too obliterated to admit of an intelligible plan. It looks like an enclosed yard, with two strongly built little chambers in one corner at the back. But the variations in the paving and the remains of internal masonry, and still more a drain which passes round three sides of a central oblong space, suggest internal divisions and a central courtyard. The floors of the little chambers are on different levels, and the higher one is traversed by flues. Possibly these chambers formed part of a small suite of baths, and the rest of the space was a house.

 p99  At Newstead, the corresponding building, which has been recently opened out, was unusually large and symmetrical. It consisted of a range of rooms and a corridor surrounding a spacious courtyard. In this, and attached to the corridor on the side next the via principalis, was a little building containing three rooms, of which the middle one was the largest and terminated with an apse. Probably these rooms were baths.

It is evident that these buildings were not on any fixed and presumably military model, but followed the lines of Romano-British houses generally, and like them had sometimes a central courtyard, and sometimes not; while some, like most rural mansions, had private baths. Their remains do not appear to have been buried in a thick mass of fallen débris, from which we may infer that their superstructures were largely of timber.


Of the large number of buildings in the forts which may conveniently be termed 'long buildings,' many were certainly the soldiers' quarters. We turn first to Gellygaer for elucidation, for the marked difference of six of its 'long buildings' from the rest promises some fruitful results.24 The buildings referred to — four in the praetentura and two in the retentura — were somewhat L‑shaped, and they ranged from 144 to 146 ft. in length, from 35 to 36 ft. across the 'head,' and about 30 ft. across the main limb (Fig. 33). In each, about two-thirds of the head was cut off by a cross-wall, and in one the space thus cut off was subdivided by a thinner section. The exploration proved that in front of the set-back wall of the limb there had been a row of timber posts, the supports of a portico or verandah. The 'heads' were next the rampart, as also were the backs of the outermost of the four and of the two at the opposite end of the fort. The group of four were in two pairs, the fronts of which faced one another from opposite sides of an intervening street.

[image ALT: missingALT]
Fig. 33. — L‑shaped Barrack-blocks at Chesters and Gellygaer
(50 ft. to 1 in.)

Chesters throws light upon these L‑shaped buildings. A pair of them, similarly placed affronté and with their 'heads' to the rampart, is partially uncovered; they are, however, more  p100 strongly constructed than at Gellygaer, and had stone columns instead of timber posts. But their most important feature is their division into a number of rooms by stone walls — the 'heads' into several of different sizes, and the limbs into series of equal size, with doors to the porticoes. The plans of these buildings, in Fig. 33, are completed by broken lines. The only other forts in Britain in which they have been found are Camelon25 and Newstead.26 In the former, two of its three were paired as at Gellygaer, and the heads were also towards the rampart; but whether this fort contained more of these L‑shaped buildings is uncertain. At Newstead, the long buildings recently uncovered in the praetentura are represented by rows of huts;  p101 but one row has a terminal longer hut which is next the rampart, and this gives it an L‑shape. We may reasonably conclude from Chesters and Newstead that the corresponding buildings at Gellygaer and Camelon were divided into a series of rooms, only by timber partitions which have disappeared.

These buildings recall the arrangement of the tents in the Hyginan camp. There, to each century, which at the time consisted of eighty men, with a centurion and petty officers, was allotted a row of tents — ten for the men, and two, or a space equal to two, to the officers — the total length of the row being 120 ft. Usually two of these rows were placed face to face, with a space between, the whole forming a striga; while a single row constituted a hemistrigium. In our plan of the Chesters L‑shaped buildings the limbs are divided into ten rooms each, while the heads, approximate equal to two of these rooms, are primarily divided by a cross-wall (corresponding with that in the single divided head at Gellygaer) into an inner single apartment, and an outer group, containing two rooms in the one case, and apparently three in the other. The Gellygaer buildings are susceptible of division into a similar number of rooms. We do not know how many of these L‑shaped buildings there were at Chesters; but there were six at Gellygaer — two strigae and two hemistrigia — that is, one for each century of an ordinary cohort.

The barracks, however, were not usually L‑shaped. At Housesteads there were thirteen simple 'long' buildings, and at Birrens a larger number. Of the former, Prof. Bosanquet27 points out that two differed from the rest, the one being a buttressed enclosure, containing at one end a small suite of baths, and the other divided into three smithies; and to these may be added a third, in consequence of its isolation and internal peculiarities. The remaining ten buildings have a family likeness. Each was divided into a number of narrow apartments, but it is not easy to determine how many, as their plans are confused by alterations and rebuildings; but eleven or twelve can be distinguished in most, and these are sometimes subdivided by a cross-wall into two rooms each. The garrison was the First Cohort of Tungrians, one of those entitled miliaria, nominally a thousand strong, and consisting of ten centuries, requiring, of course, ten barracks.  p102 Unfortunate, the plans of these buildings at Birrens are less perfect, but their larger number may be feasibly explained by the fact that the garrison — the Second Cohort of Tungrians — was not only miliaria but equitata, that is, it included a detachment of horsemen, for whose use stabling and other accessories would be required, in addition to barracks. It is probable that the largest of the three storehouses was for the storage of the horses' forage, and that the long buildings of the praetentura, in which it was situated, were the quarters of the cavalry and their horses.

The long buildings in the praetentura at Newstead were, as already stated, represented by rows of huts. Each row was about 190 ft. long, and contained eleven of these huts, all of similar size, with the exception above referred to, and averaging 35 by 15 ft. Their foundations were very slight, indicating that their superstructures were probably of timber. There were twelve of these rows, four of them single, or hemistrigia, and the rest paired, or strigae, the whole forming two groups, one on either side of the longitudinal street and the hemistrigia were outermost. The foundations of the long buildings in the retentura show that they were of the ordinary type, and most have no remains of cross-walls dividing them into rooms. The barrack accommodation thus was greater than that of Housesteads or Birrens, but it was a much larger fort, with presumably a larger garrison.

A number of L‑shaped barracks have been brought to light on the site of the legionary fortress at Neuss on the Continent, the head-quarters of the Sixth Legion before its removal to Britain under Hadrian. "The normal barrack block at Neuss," to quote Prof. Bosanquet, "measures 240 by 80 ft., and consists of two long buildings face to face, separated by a road 16 ft. wide. Each building is in two parts. At the end nearest the rampart was a wing, 80  ft. long, divided into numerous rooms and mess-rooms. The remainder, a strip 165 ft. long, is divided into twelve exactly similar compartments, each consisting of a front and a back room. Their front wall is considerably behind that of the 'mess-house wing,' leaving room for a verandah 9½ ft. wide." This division of the compartments into a front and a back room recalls the similar division at Housesteads, and  p103 the compartments at Chesters being long and narrow are suggestive of a similar subdivision, probably by timber partitions.

The barracks appear to have been of a single storey, and as a rule of comparatively slight construction. At Ardoch, they were wholly of timber. At Gellygaer, they had stone foundations, but the small amount of débris on their sites led the explorers to consider that these foundations carried timber structures. The areas that would be occupied by them at Hardknott and Melandra have yielded no stone foundations, but in the latter fort, clay floors and oak posts have been found. Those at Chesters may have been of stone; and Prof. Bosanquet considered that the final ones at Housesteads were also of stone, but were preceded by timber structures on stone foundations. No roofing-tiles have been reported as found on their sites, and from this may be inferred that they were covered with thatch or wood.

Other Buildings and Structures

On all the more complete plans of our Roman forts may be distinguished large and important buildings, which from their situations, internal divisions, and other peculiarities cannot be classed with any of those already described, and with few exceptions it is impossible at present to assign to them their uses. At Gellygaer, five of these 'other buildings,' of different shapes,28 are readily distinguished, all adjacent to the central zone of principal buildings. One of these behind the commandant's house was divided into three large compartments, which, to judge from the objects found in them, were workshops. The irregular block, which balanced it on the other side of the longitudinal street, had probably a like use, as it lay at the back of a large enclosed yard, which yielded evidences of industrial pursuits. This yard was between the head-quarters and the storehouse on that side of the fort, and was entered by a wide gateway from the via principalis. Nothing was found in the three blocks along the opposite side of the via principalis to throw any light upon their uses.

It was mentioned above that two, if not three, of the long buildings at Housesteads differed from the normal barrack-blocks,  p104 one being a buttressed enclosure with bathrooms at the end. At Birrens, there was a similar but larger buttressed enclosure on one side of the headquarters, and external to it along one side the remains of bathrooms, the whole forming a square block.29 As already observed, the commandants' houses at Chesters, Camelon, Great Chesters, Newstead, and Rough Castle appear to have contained baths. There were two such houses at Housesteads and at Birrens, but without indications of baths. It is not unlikely, therefore, that in each of these forts these houses were served by a detached bathing establishment. At Ardoch, also, the remains of hypocausts and flues near the north gate may have related to detached baths.30

Of the arrangements for the preparation and cooking of the food of the garrisons we know practically nothing. At Birrens, the lower courses of a row of four oven-like structures,31 blackened with charcoal, were found on the inner side of the rampart near the east gate. Each was about 5 ft. 6 ins. in diameter, with a paved floor, and a narrow opening to the intervallum. It was reasonably conjectured that they were the remains of ovens for the cooking of food. At Housesteads, one of the storehouses contains in the middle the remains of a similar circular chamber, with a narrow opening or flue through the south wall of the building; and the east guard-room of the south gateway contains another.32 The opening of this is carried through the south wall into an external building, 27 ft. by 21 ft., erected against the guard-room. When first explored, there was evidence of a stone floor covered with cement above the 'kiln' in the guard-room, also hard by was an oven with a vaulted roof. All these remains at Housesteads appear to be late work, but there is no evidence that they are post-Roman, for at Birdoswald there were formerly the remains of two similar oven-like structures, about 4 ft. 4 ins. in diameter, one on each side of the south gate; and at Great Chesters a similar structure just within the rampart and close by the east guard-room was found in 1894, and there is evidence that a century or more ago there was a corresponding one outside the west guard-room.33 A similar structure was found within the north-east corner at Haltwhistle. The Housesteads examples  p105 are generally regarded as kilns for drying wheat or malt, and certainly their deep form with sloping sides renders it unlikely that they were ovens.34

The construction of the roads or streets of the forts has nothing to distinguish it from that of Roman roads generally. At Housesteads, the surface of the streets was here and there the bare rock; elsewhere it was of masons' chippings on a rough pitched foundation, or a pavement of cobbles or flags on a basis of chippings. At Birrens, one of the streets is described as a thick consolidated bed of gravel with a raised crown, retained by kerbs of large flat stones, while other streets were paved with cobbles on a bed of gravel. At Gellygaer, Camelon, and Rough Castle they were mostly of gravel on a bottoming of broken stone. At Cardiff Castle, the roadway through the north gateway was of iron scoriae on a bed of rough stones; and below this were two older roadways of the same construction.35

The drainage of the forts was well considered and carried out. At Gellygaer,36 in spite of the ruinous condition of many of the drains, the general system could be ascertained. The ground falls to the south-south‑east, and this determined the main lines of the drains. Two arteries were found, the one passing through the south-west gateway and debouching into the ditch there, but with a bend to avoid the bridge; the other traversing the whole length of the via principalis and similarly making its exit through the south-east gateway. The main drains were from 1 ft. 6 ins. to 1 ft. 10 ins. wide, and from 1 ft. 6 ins. to 2 ft. 4 ins. deep, and appear to have all been covered with large flagstones from 4 to 6 ins. below the surface of the roads. The sides were carefully built in regular courses, and the bottoms were paved or pitched or were the natural clay of the site. The smaller drains were, as a rule, similar; but one was neatly constructed of thin slabs set on edge and roofed with similar slabs. Along the sides of some of the streets the drain was simply the space between the kerb and the foundations of the adjacent buildings, but it is not clear whether these were covered or were open water-channels. These underground drains at Gellygaer may be accepted as generally representative of those of other forts. Small drains were occasionally V‑shaped, the sides and tops being of flags.

 p106  Several street water-channels or gutters were found at Birrens. The chief streets appear to have had one on either side, and the smaller a single one along the middle. These gutters took the form of rectangular channels about 9 ins. wide and 4 ins. deep, cut in long blocks of stone about 18 ins. wide, which in the chief streets were placed outside the kerbs. The outlet of one of these gutters remained. The water passed under a large flat stone on the road-level, which covered a hopper with built sides and a sloping bottom leading down to the drain.37

A plentiful supply of water for drinking and cleansing purposes was one of the first considerations. We have noted the presence of a well in some of the courtyards of the headquarters buildings.38 But one or even several wells would hardly meet the varied demands for water, even in a small fort. For a more abundant supply, recourse was had to external springs or streams. At Great Chesters,39 water was conveyed from Haltwhistle Burn by an aqueduct 5 miles long, in the form of a small canal, 3 or 4 ft. wide, cut in the sides of the intervening hills. At Birdoswald,40 the water of a spring several hundreds of yards distant was conveyed by a culvert constructed of large slabs of stone to a cistern near the centre of the fort. At South Shields and Chester, inscribed slabs have been found recording the construction of aqueducts under the direction of Morius Valerianus and of Alpius Marcellus respectively.41 This is specially interesting in the case of Chesters, as not only had that fort an unusually large well, but, close by, the North Tyne provided an inexhaustible supply of pure river-water.

Open cisterns or tanks have been found in several forts, notably at Housesteads, the largest of which, near the south-east corner, is nearly 15 by 10 ft., with sides 3 ft. high.42 The sides are of ten large slabs of stone, kept in position by tongues of lead, and in addition by iron clamps on the top and iron stay-bars let into the cement floor below. A stone coping ran round the top, of which two lengths remain. As no trace of an inlet is to be seen, the supply may have poured over the top through a lead pipe. The surplus water escaped through a  p107 notch at the west end, and was conveyed by a stone gutter to a latrine close by; and a hole for a tap or plug half-way down at this end provided an occasional means of more thoroughly flushing this latrine. There were at least four other smaller cisterns, of similar but simpler construction, in this fort. At Gellygaer, the remains of a large cistern, 20 ft. by 7 ft. 6 ins., were found, facing the head-quarters and attached to the end of a long building.43 Sufficient of the sides remained to show that they had been built of tile, and the bottom was of brick concrete, on a foundation of two layers of pitched stones, while external to all was a pugging of clay. At one end of the cistern was a lead outlet pipe to the adjacent drain. There was no indication how this cistern was supplied.

As one of the first aims of an enterprising enemy would be to cut off the external water-supply — and how easily could the five-mile aqueduct at Great Chesters, the whole length of which lay on the Caledonian side of the wall, have been thus rendered useless! — we can well imagine that at least one reliable well within the ramparts would be regarded as absolutely necessary. Apparently the normal position of a well was within the precincts of the headquarters, so that in times of stress its limited supply would be under direct official supervision and regulation.

Remains of latrines have been found at Housesteads, Castlecary, Gellygaer and Bar Hill. The latrine of the first of these44 was erected against the south wall of the fort, near the cistern above described, and in a situation well adapted to receive the surface-water to flush it, the ground here sloping to the south. This building was 31 by 16 ft. internally, with a door at either end, but the eastern, which was blocked, was probably the original one. Along the two sides and the western end of the interior was a continuous trough, 3 ft. wide and 2 ft. 6 ins. deep, with a flagged bottom and built sides, the outer side being the wall of the building. The intervening floor was paved and had a marginal gutter. This gutter received the surplus water of the cistern, as described above; and its gentle fall from the south-east to its north-east extremity allowed every portion of it to be flushed, and the water finally poured into the trough at the latter extremity, where it also received the street drainage. Presumably the trough bottom sloped in a contrary direction,  p108 with an outlet through the fort wall at the south-east end. As the small baths, referred to on page 101, were due north, it is likely that their waste water also contributed to the necessary flush. "Above the trough, seats were doubtless arranged in the same way as at Uriconium, but there was no visible provision for the woodwork beyond a rebate formed on top of the inner trough wall, which may have supported a sill-piece."

The latrine at Castlecary45 occupied a similar position near the north-east corner of the fort, but was on a smaller scale, the internal dimensions being 16 by 12 ft. It was similarly built against the fort wall, and it had a single doorway at the west end. The flush appears to have been wholly derived from the neighbouring street drains, especially from one from the south. The latrine at Gellygaer46 was within the enclosed yard, near its north corner, but it was in too ruinous a condition for its whole plan to be made out. The latrine at Bar Hill47 was at the lower end of a narrow bath range extending from the south gate to the south-west angle of the fort, but the whole range was in an extremely ruinous condition.

'Suburban' Buildings

In Chapter II, some of the castella were described as having fortified enclosures or annexes attached to them, and incidental reference was made to external buildings, notably large baths. Unfortunately little is known of the 'suburbs' of the forts, but the indications of buildings in their vicinity is by no means uncommon. Housesteads is a notable example. To the south and east of this fort are the foundations of streets and buildings which a century or more ago were so conspicuous that old writers described the site as that of a city. Altars, statues, columns, and carved stones have been turned up from time to time, and they tell of temples (the remains of two of which are known), baths, and other goodly structures. There is no doubt that the suburbs of Borcovicus sheltered a considerable population. In the vicinity of other Wall forts may be discerned the indications of former occupancy. Hard by Chesters and Great Chesters  p109 are the remains of baths, on a large scale, and similar remains may be seen or have been revealed by the spade close by the Roman forts at Camelon, Inchtuthill,48 Rough Castle, Halton Chesters, Chester-le‑Street, Binchester, Slack, Caerhûn49 near Conway, Gellygaer, and Caersws, while at Lanchester, Plumpton, near Carlisle, Binchester, and Old Carlisle, are the remains of large buildings, which probably also were baths.

These large bath buildings were unquestionably for the use of the garrisons, and must be distinguished from the small baths within some of the forts. They were placed without the walls, probably on account of the large spaces they occupied; perhaps also for the convenience of the suburban inhabitants. The normal situation was within the annexe, when there was one: it was so at Camelon, Rough Castle and Gellygaer. Not every fort had an annexe. Not one has been reported along the line of the Wall of Hadrian. Lyne had two wing-like annexes, and Camelon also two, one larger than the castellum itself. Little as yet is known of the contents of these fortified enclosures. The via principalis at Camelon was continued through its larger annexe and was crossed by another street; and besides the baths there have been found within, the remains of another large building and traces of others. The present exploration of the Gellygaer annexe has proved that it was traversed by the via principalis, and besides the baths there was a large yard attached to them and containing remains of furnaces; while on the other side of the street have been found another large enclosure with a wing, and other structures.

In the suburbs of a fort should be expected the dwellings of the families of the soldiers, besides those of other civilians who served the garrison in various capacities, or who elected to live under the shadow of the stronghold. The religious needs of the community would be met by shrines and temples as at Housesteads. So far the excavations at Gellygaer have not revealed these various structures, and there does not appear to have been room for them in the annexe. Probably the spade will yet reveal traces of a civil settlement elsewhere.

The Author's Notes:

1 Roman Wall, p348.

2 Soc. Antiquaries Scot. xxx, p110.

3 Roman Fort of Gellygaer, p51.

4 Arch. Aeliana, XXV, p208.

5 Cumb. and West. Arch. Soc. xii, p406.

6 Melandra Castle, p56.

7 Arch. Aeliana, XXIV, p55.

8 Derbyshire Arch. and N. H. Soc. xxvi, p186.

9 Soc. Antiq. Scot. xxxix.

10 Ib. xxxv, p364.

11 In the vault under the sacellum at High Rochester was found among the débris an altar dedicated to the Genius of the Emperor and the standards — Roman Wall, p318; and an altar dedicated to the Discipline of the Emperor, found in the headquarters well at Birrens, possibly came from the sacellum. See Roman Era in Britain, p131.

12 See Figs. 64 and 65.

13 Arch. Aeliana 3rd ser. v.

14 An inscription at Great Chesters records the restoration of a horreum, and on the site of one of the Corbridge buildings was found an altar dedicated by the praepositus of the horreum.

15 The east granary at Corbridge contained a row of pillars, and this suggests an upper floor; but the pillars did not form part of the original construction, and may have been introduced to support a defective roof.

16 Roman Fort of Gellygaer, p59.

17 Soc. Antiquaries Scot. xxxv, p180.

18 Arch. Aelian. xxv, p239.

19 Soc. Antiq. Scot. xxx, p112.

20 Arch. Aeliana, III (O.S.), p142.

21 Ib. xxiv, p56.

22 Soc. Antiquaries Scot. xxxv, p363.

23 Ib. xxxix p39.

24 Roman Fort of Gellygaer, p65.

25 Soc. Antiquaries Scot. xxxv, p362.

26 Interim Report, 1907.

27 For his valuable observations on 'barracks' see Arch. Aeliana, XXV, p228.

28 Roman Fort of Gellygaer, pp67‑71.

29 Soc. Antiquaries Scot. xxx, p112.

30 Ib. xxii, p449.

31 Ib. xxx, p101.

32 Arch. Aeliana, XXV, p236.

33 Ib. xxiv, p37.

34 Probably most of these oven-like structures contained cauldrons or boilers.

35 Archaeologia, LVII, p343.

36 Roman Fort of Gellygaer, p30.

37 Soc. Antiquaries Scot. xxx, p117.

38 Two were found at Birrens, one to supply the baths.

39 Roman Wall, p234.

40 Ib. p261; also 349.

41 Proc. Soc. Ant. 2, xiv, p293; xvi, p387.

42 Arch. Aeliana, XXV, p248.

43 Roman Fort of Gellygaer, p69.

44 Arch. Aeliana, XXV, p249.

45 Soc. Antiquaries Scot. xxxvii, p50.

46 Roman Fort of Gellygaer, p71.

47 Roman Forts on the Bar Hill, p44.

48 Proc. Soc. Ant. Scot. XXXVI, p224.

49 Archaeologia, XXVI, p127.

[image ALT: Valid HTML 4.01.]

Page updated: 3 Mar 18