[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 10
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 12

 p254  Chapter XI


Walls, Roofs, Floors, Doors, and Windows, Heating of Rooms, Water‑Supply, and Treatment of Internal Walls

The work of the Roman builder is more easily distinguished in this than in other countries, for, with the passing of the empire, building became well-nigh a lost art with us for several centuries. On the Continent, and especially in Italy, Roman art survived, but under the new conditions it rapidly changed, and thus changed was reintroduced into Britain under the later Saxon kings — the advent of an old friend in new guise, for Roman had become Romanesque. Out of this developed the graceful pointed styles of the medieval period, when the builder's art attained its zenith in this country. The whole spirit of architecture was now transformed. Columns, arches, mouldings, and sculpture showed little of their Roman parentage, and not much experience is needed to distinguish their fragments from those of Roman work. With the Renaissance which followed, the case is different. Roman forms and decoration were more or less closely copied, and stray fragments of these are liable to be mistaken for their ancient prototypes. At first Roman art was freely adapted to modern requirements, but this gave place in the eighteenth century to a servile copying of classical models, which resulted in temple-like buildings singularly inappropriate under our grey skies. The proportions of columns and entablatures, however, were less founded upon Greek and Roman examples, than upon the rules laid down by Vitruvius and other writers. Even in ancient Greece and Italy the proportions did not slavishly follow set rules; still less in Roman Britain, where much of the architecture was a crude  p255 edition of the current architecture of Italy. But even in this remote land the designs and technique sometimes vied with the best productions of nearer provinces to the seat of the empire. While the Roman work in this country appeals to us as the product of a people who better excelled in engineering and constructive skill than in art, it must not be overlooked that the most conspicuous and best-known remains are military, where utility would be the first consideration. The public buildings, temples, and larger houses, where architectural effect might reasonably be looked for, are now little more than crumbling foundations and fallen materials; and these occurred, as a rule, in regions where stone suitable for carving is scarce or unobtainable. There is evidence of the general use of stucco, which has crumbled, and of timber, which has wholly disappeared. But more is known of the interiors; and from the rich mosaic pavements and the fragments of painted mural decorations, which have been abundantly found, it can hardly be thought that their exteriors would lack embellishment. The remains of the forum and basilica at Silchester indicate an ensemble of buildings at once spacious and stately; and the houses, however plain their exteriors, must have had a certain picturesqueness arising from their outshoots and corridors, and roof-lines of different heights.


The reader will have already observed that the masonry of the era varies considerably, and frequently has little or nothing to distinguish it from that of later times. The Pennant-grit walls of the Roman fort at Gellygaer are precisely similar in their appearance and method of construction to the modern work in the same material in that district; and it would be a difficult task to point out wherein the walls of the Roman houses at Silchester differ from the flint walls of later date. The Roman builders, like the modern, were governed by the material they worked with. Still, their masonry has certain peculiarities which broadly distinguish it. The faces of the stones are often short or squarish, and this is owing to a predilection for 'headers' rather than for 'stretchers.' There is also a marked tendency for the stones to be narrowed off behind, so as to firmly key into  p256 the core. The use of mortar mixed with broken brick in exposed positions or where great strength was required, is also characteristic. In the south-east of the island, where only flints and other small stones were available, the lacing-courses of tiles to secure the thin facings to the core are highly characteristic of Roman work.

The masonry was almost invariably in regular courses. The facing-stones were nearly always squared, if only roughly so; and even in the coarsest work the faces had their irregular projections removed by the hammer. At Gellygaer, the punch was here and there brought into requisition to finish off the work of the hammer. Sometimes, and especially in the north, the faces were wrought in a peculiar manner, the tooling taking the forms of reticulated zigzag and feather-like scorings ½ in. or more apart. In more finished work, as in the jambs of gateways and the like, the stones were carefully dressed with the punch, as in Fig. 72, or the chisel. In Fig. 73, another Corbridge example, the rough-pitched faces have chisel-drafted margins, and the joints closely fit without mortar. This work is confined to masonry of large stones, and other examples may be seen at Housesteads. Random work is rare. An example of this was recently uncovered at Cwmbrwyn,1 Carmarthenshire, where the stones were used in almost the rough condition in which they were quarried.

[image ALT: zzz]

[image ALT: zzz]

In ordinary walling the joints were usually wide, and the mortar coarse and often mixed with small gravel. In spite of the proverbial excellence of Roman mortar, it did not always attain a high standard in Britain. Now and again it was little better than a mixture of earth and lime, and the latter has sometimes been removed by the solvent action of the moisture of the soil, leaving an earthy residue which has misled observers into thinking that puddled earth was used for mortar. In narrow walls, as those of houses, and occasionally in the thick, defensive walls of cities and forts, as at Caerwent, the mortar was wholly trowel-laid; but more often the latter were grouted internally, as already described in the case of Cardiff (p52). The pointing at these two places is here and there in a remarkably fresh condition. At Cardiff, it seems to be of the mortar used for bedding the facing stones, which is finer than usual, and  p257 was probably done as the work proceeded. At Caerwent, a white mortar was used that contrasts in its fineness with the gravelly mortar of the interior. In both, the pointing would be described by a modern mason as 'flush-pointing.' In some work recently exposed at Caerleon, the masonry was dressed and the joints filled with fine mortar, on which sunk bastard joints were ruled with a rounded tool, and then painted red. At Burgh Castle, pink or fine brick mortar was used for the pointing. These brick mortars, whether fine or coarse, have generally stood the test of time well, and they were much used in the construction and lining of hypocausts and baths, and generally where strength and tenacity were required. Clay was frequently used instead of mortar in foundations and revetments, in furnaces and hypocausts, where the heat soon gave it a bricklike consistency, and to render tanks watertight. It is probable that the fine, hard, white cements for lining the floors and walls of rooms were mixed with the whites of eggs or saccharine fluids.

Walls built of, or faced entirely with, brick were frequent in Rome at a late period, but were exceptional in this country, the use of brick, or, as they are usually called, tiles, being generally confined to the construction of hypocausts and furnaces, and to form lacing-courses. Arches were frequently turned in them, and sometimes they served as quoins in flint and other walls constructed of small stones.

Roman bricks are almost invariably of red clay. The kneaded clay appears to have been first rolled out on sand to the desired thickness, and then cut into suitable sizes to be transferred into shallow frames or moulds, in which they were pressed out. When removed, the soft pieces were finally scraped or trimmed with knives. Four sizes are commonly met with in this country, three of them square, approximately 7¾ by 7¾ ins., 11 by 11 ins., and 16¼ by 16¼ ins., and an oblong, approximately 16¼ by 11 ins.; and less frequently a larger square, about 23 by 23 ins. These are the average dimensions: actual specimens may vary from ½ to ¾ in. in the smaller, and 1 in. or more in the larger sizes. The differences are probably due to unequal shrinkage in the manufacture, some clays shrinking more than others, and even the same clay, according to its moistness or the degree of heat it is exposed to. The smallest size mentioned above was made specially for the pilae, and the largest for the roofs of the hypocausts.  p258 The two medium square sizes were evidently intended for a Roman foot and foot-and‑a‑half respectively, and so correspond with the tegulae pedales and tegulae sesquipedales of ancient writers. The oblong bricks could be used with either of these according to the way they were placed, and probably were made for effecting a bond with them. Intermediate sizes of rectangular brick are decidedly rare. Half-round bricks are occasionally met with, and were probably used, as in Pompeii, in the construction of pillars. Roman bricks rarely exceed the limits of 1½ and 2 in. in thickness.

Few Roman arches, even in an imperfect condition, remain in this country. They appear to have been almost invariably semicircular and of a single ring of voussoirs but flat arches are known to have existed. Shaped voussoirs were used in the large arches of the gates of forts and cities. At Gellygaer and Merthyr Tydfil they were of a local calcareous tufa, which was selected on account of the ease with which it could be sawn. Brick voussoirs were common. A large fallen piece of a brick arch, which crowned a lunate window at the west end of the great bath at Bath, is to be seen there. Small arches, however, were often constructed of ordinary bricks or of thin flat stones, in which case the joints were necessarily wedge-shaped, and the stability of the structure depended upon the strength and cohesiveness of the mortar. For vaultings, where lightness was a first consideration, hollow brick voussoirs were used. Those of the main vault of the great bath just referred to, were 1 ft. long by 4 ins. in depth, and tapered on their faces from about 6½ to 7½ ins. They were scored with a comb-like instrument or 'scratch,' in order to make the mortar adhere to them. The upper surface of the vault was covered with a thick layer of coarse mortar, into which had been pressed half-round roofing-tiles (imbrices) arranged longitudinally so as to lace together the successive zones of voussoirs.2 The under surface of the vault  p259 was lined with fine white mortar, which may have been ornamented, as in Pompeian examples, with raised mouldings and arabesques. In Fig. 74 are shown two hollow voussoirs from Bath, the second being remarkable for the puzzling semicircular notches near its base.

[image ALT: zzz]
Fig. 74. — Tile Voussoirs, Bath. (⅛.)

That timber was extensively used in the construction of houses can hardly be questioned, but direct evidence is scanty. Again and again at Silchester and elsewhere, undisturbed floors of small buildings have been found without a trace of the walls that enclosed them, and the presumption is that these were of timber. Internal timber partitions are indicated by the chases in the floors for their sleepers, or plinths.3 Rows of post-holes marked the sites of timber buildings within the fort at Ardoch (p30). Evidence of a positive nature was supplied by the remains of a house discovered at Bucklersbury, London, in 1869.4 Here all the walls, with the exception of the apse of a large room, had been of timber-work, and fragments of the oak sleepers remained in situ.

Almost invariably, however, the houses — especially the larger ones — had stone foundations for both external and internal walls; but this is no proof that their superstructures were of masonry. Our existing old timbered houses have masonry plinths or basements, and there is good reason to think that this was customary in Roman times. The sites of the Roman houses are not as a rule so encumbered with fallen building stones as might have been expected had their walls been wholly of masonry; and now and again the foundations are too narrow and slight to have borne such weighty structures. Mr. St. John Hope, in describing a house uncovered at Silchester in 1901, sums up his experience as follows: "For my own part, I have long been convinced that a large number of the buildings at Silchester were of half-timber construction, and that only in the case of rooms warmed by hypocausts, which are usually placed at the ends or sides of the houses, were the walls carried up to their full height in masonry to avoid risk of fire from the flues built into them. This is also borne out by the greater thickness of the walls of these winter rooms."5

 p260  This house (1, XXVII) furnished some interesting evidence as to the construction of the Silchester houses. It, or a portion of it, had been destroyed by fire, and amongst the burnt débris were pieces of clay parget, baked into brick. On the front had been impressed a herringbone pattern from a wooden stamp; and on the back were the marks of wattlework. The inference is clear enough. Some portion of the house — probably the upper storey — had been of timber framework, with the panels filled in with wattle and daubed over with clay — a method of construction which was common enough up to two centuries ago, and which is still to be seen in old cottages, although the 'wattle-and‑daub' has usually been replaced by brickwork. That this was not exceptional at Silchester is proved by the frequent presence of a layer of pasty clay overlying the floors of the houses, which had long puzzled the explorers. The impress of wattles or laths has been noticed on pieces of the internal wall plaster of Roman houses, notably at Caerwent.6

Nails are among the most frequent 'finds' on Roman sites, but, as might be expected, they are, as a rule, reduced to shapeless masses of rust. In Fig. 75 are shown the prevailing forms. They appear to have always been square in section, and by far the commonest forms had tabular heads, square or circular, as  p262 the first four in the illustration. That in the middle resembles a modern 'clasp' nail, and the remaining ones, except the small one with a hemispherical head, are variants of it. As a whole, they resemble the old-fashioned hand-made, before the introduction of machine-made ones.

[image ALT: zzz]
Fig. 75. — Roman Carpenters' Nails. (⅓.)



[image ALT: zzz]
Fig. 76. — Tegula and Imbrex (⅛) and their combination on a roof.
On the sites of most Roman buildings are found roofing-tiles and flags, showing that roofs so covered were common in Roman Britain. The tiles were of red clay, and of a form which still prevails in Italy and elsewhere in southern Europe. They were in two series, which were as necessary to one another as the warp and woof of a woven fabric — flat tegulae with upturned edges and half-round imbrices, A and B, Fig. 76. The tegulae were roughly 1 ft. 10 ins. long and 1 ft. 4 ins. wide; but these dimensions varied, owing probably more to shrinkage in the process of manufacture than to intentional differences. They were usually about 1 in. thick, and the flanges were as wide and high. The flanges of each tiles stopped short of the upper end by about 2½ ins., and the lower angles were usually notched out on the under side to receive the upper ends of the flanges of the tile below in the row, in order that the tiles should be well locked together and have the necessary overlap. To facilitate this still further, the tiles were sometimes slightly wedge-shaped, the lower ends being narrower than the upper. The imbrices were  p263 about as long as the tegulae, and they were tapered, so that the wider lower end of each enclosed the upper end of its fellow below. These covered the flanges of the flat tiles and were secured by mortar, the roof thus presenting a series of bold rolls with intervening flat channels, as shown in Fig. 76. As these tiles were not provided with nail-holes, they were only suitable for roofs of low pitch where they could remain in position by their own dead weight, and with little doubt the roofs were strongly planked, not lathed, to receive them.

The roofing-tiles, like the bricks, were made by hand, aided perhaps by frames or moulds to give them their requisite general form, and the imbrices were probably curved over shaped wooden blocks. By way of decorative effect, concentric groovings were often produced on the lower portions of the tegulae by a rapid semicircular sweep of a tool, or possibly of the fingers.

Antefixes for ornamenting the eaves of these roofs (Fig. 77) were only sparingly used in this country. Examples have been found at Chester, Caerleon and a few other places. The tongues  p264 behind were pushed into the ends of the lowest imbrices and secured by mortar.

[image ALT: zzz]
Fig. 77. — Antefixes, Chester and London. (¼.)

The slabs or 'slates' of the stone roofs were of a peculiar and characteristic form. The sides were parallel, or slightly converged upwards, and instead of being square at the foot, as is usual at present, they were pointed. The upper ends, when shaped at all, were roughly pointed, thus giving the whole slab an elongated hexagonal form. The usual width was about 11 ins., and length from 16 to 18 ins. or more. These slabs were invariably held by nails, and as the distance of the nail-hole from the point varies considerably even in the slabs on the same site, it is almost certain that they were fastened to boarded roofs. In combination they gave rise to a lozengy pattern, as shown in Fig. 78. Those used at the eaves were square at the base, and those next the ridge were of the ordinary form but shorter and with flat heads. The material of the slabs depended, of course, upon what was available, as the fissile limestone known as Stonefield slates in the south-west of the island, and Pennant flags in Monmouthsire and South Wales, while true slates were used in Cumberland, and at Cwmbrwyn in Carmarthenshire, an ashy slate from the Precelly Hills.

[image ALT: zzz]
Fig. 78. — Stone Roof with Cresting and Finial, Llantwit Major.

The ridges of the tiled roofs were probably crested with ridge-tiles resembling the imbrices, if not with imbrices themselves. But many examples of the stone crestings of the stone and slate roofs have been found. On the site of a house at Llantwit Major in Glamorgan,7 the Pennant slates of the roof lay as they fell, and with them lengths of the Bath stone cresting. These had flat tops, 6 ins. wide, and sloping sides of about 4 ins., while the under sides were roughly channelled to receive the ridge, and retained traces of mortar. With them lay a gable-finial of Bath stone, the base of which was similarly channelled below and formed the terminal of the ridge-cresting. Its oblong body was perforated with two intersecting arched openings, thus leaving four square piers, each with a simple cap, and the summit feature was a foliated knob. Similar finials have been found at Bath and elsewhere in the west of England. The Llantwit Major roof, with its lozenge-shaped slabs, bold cresting, and graceful finials, must have had a pretty appearance, and might well be copied in modern work.

 p265  Roof-coverings of less durable materials were certainly frequent in Roman Britain, as it is not unusual to find neither tiles nor slates on the sites of buildings. At Gellygaer, for instance, the barracks and other long buildings contrasted with the central buildings in the absence of roofing-tiles on their sites, the inference being that they were covered with wood or thatch.


Timber must have been exceptional in Roman Britain for ground floors, for their sites nearly always yield traces of flooring materials of a more enduring nature. The absence of these traces is, of course, no proof for timber; but occasionally the soil on the site is not only devoid of them, but is in a soft and natural condition, while contiguous rooms have mortar or mosaic pavements, and in such a case it is reasonable to infer a timber floor. But there is evidence of a more direct nature. At Silchester, a patch of gravel in Insula XVI had a series of parallel trenches about 6 ins. wide and 18 ins. apart, and filled with dark mould, implying a series of joists which had decayed.8 And in several instances the walls of houses and other buildings there exhibited on their inner sides flat off-sets apparently for beams to carry floor-joists. Upper floors can hardly have been otherwise than of timber, and apparently wholly so, for nothing has been recorded that could be construed as fallen mosaic or plaster from floors above.

Floors of beaten earth, gravel, or clay were mostly confined to outbuildings and cottages, or to rooms of little importance.

Mortar and concrete floors were very frequent, and are found on the sites of most houses for kitchens and other rooms used for menial purposes, and sometimes even for the best rooms. They varied in construction and finish, from a mere spread of coarse mortar or concrete upon the natural soil or upon a bed of broken stone, to a layer of fine mortar with a surfacing of fine lime-cement, resting on a foundation of concrete. These fine white floors must have vied with the mosaic pavements in beauty. One found in House 1, VIII, Silchester, had a fine brick concrete basis, finished off with a coat of white-lime cement, which when  p266 polished must have had an ivory-like surface, contrasting with the gay colouring of the walls.9 The quarter-round skirting of this room was in a fine pink cement. Floors of this class are rarely perfect, and it is likely that many which have been described as of indurated gravel, rammed chalk, or loamy sand may have originally been of concrete or mortar that has partially or completely lost its limy constituent.

Opus signinum floors properly come under this head, but the term is often applied to the brick concretes which the Romans used for a variety of purposes besides floors, and more often as the basis for cement or mosaic than as the actual floor-surface. The term should be confined to the more carefully made concretes of this sort which were ground and polished, giving rise to an artificial polished breccia. In order to give an enriched effect to these floors, chippings of marble and other fine stones were occasionally introduced, as in the floor of one of the temples at Silchester.10

Flagged paving and pitching were chiefly used for stables and outbuildings, and especially for yards and other spaces open to the sky. The schola of the great bath of Bath was paved with huge slabs of freestone, well squared, but of different sizes, yet well fitted together; but probably this was the foundation for a mosaic floor. At a later date, however, the bathers walked upon a similar pavement of smaller slabs at a higher level. The yards of the headquarters and of the baths at Chesters were similarly paved with large squared slabs; but more often these pavements were of less careful construction, consisting of blocks of all shapes and sizes as quarried or collected from the surface, and fitted together as best they might be. In chalk regions flints were much used for the purpose. Tile pavements are occasionally met with. In several found at Silchester the tiles were carefully laid in brick mortar. The design of one of these pavements (House 1, XXIII),11 as originally intended, consisted of a border of two rows of tiles, 8½ ins. square, the enclosed space having octagonal tiles of the same size with small square ones to fill the interspaces. The builder, however, seems to have run short of these tiles, and so completed his work with hexagonal tiles of smaller size, filling the interspaces with rough mosaic. Such  p267 pavements may be regarded as rude imitations of opus sectile, which consisted of pieces of marble or tile cut into various shapes to make up geometrical patterns.

But the pavements which are best known and most appreciated are those of mosaic. As might be expected, in a remote province they are not of the costly materials and fineness of those of Italy, yet some approach them in these respects. The advantage of mosaic lies in the facility with which small cubes of coloured materials can be worked into patterns of varied and gradated colours. The cubes ranged from ½ in. to 1½ ins., rarely more or less, and in a general way two sizes prevailed, a smaller, from ½ to ¾ in., and a larger, from 1¼ to 1½ ins.; and according to the size used, we may conveniently divide mosaic-work into 'fine' and 'coarse.' The former was used in decorated work; the latter chiefly in the plain or only slightly decorated work of corridors and less important rooms. The faces of the tesserae in the finer work were roughly square; but in the coarser they were more often somewhat oblong. The materials of the tesserae were, as a rule, tile and stone; but in the finest decorative details coloured glass was occasionally used. In a country of such diversified geological structure as ours, it was rarely necessary to go far afield for suitable natural materials. Sandstones provided buff, grey, dull red, and even purple tesserae; limestones and marbles, an even wider range of colours from white to almost black; certain ironstones, chocolate and black; and hard varieties of chalk, white. Some of the sandstones appear to have had their colours modified by heat. Tiles, of course, provided various shades of red, from salmon to purple brown.

In plain mosaics the tesserae were laid in rows running parallel to the sides of the room. As they varied in length, the rows necessarily 'broke joint' like the course of a wall. In decorated work the rows followed the lines of the pattern. The fan-shaped arrangement in modern work is never seen in the Roman.

The tesserae were bedded in fine white or pink cement of great hardness, and the greatest care was exercised to obtain a good and level foundation. In the Basilica of Viroconium12 the white setting rested upon 2½ ins. of fine brick concrete, which was spread over a foundation of broken stones levelled up with mortar  p268 2 ft. thick. One of the mosaic floors of House 1, XIV, Silchester,13 had the following in descending order: red cement, 1½ ins., upon which the tesserae, bedded in pink cement, were laid; white concrete, 3 ins.; and yellow mortar, 8 ins. Another in the same house had its tesserae similarly laid in pink cement, and resting on the following understructure: drab cement and mortar, 3½ ins.; white cement and pebbles, 3 ins.; gravel, 5 ins.; and mortar, 1 in.

There is little doubt that the plain mosaics were laid by the direct process, that is, the cubes were placed directly on the prepared floor-surface; but whether this was the process adopted in the decorated ones is less certain. The usual method at present is first to cut a number of pieces of stout paper which fit together to correspond with the intended mosaic, and on these the pattern is traced. The cubes are then glued or pasted upside down on them, and when dry are ready to be transferred to the freshly cemented surface. The sections are applied with the cubes down wards, and, after being well pressed into the soft cement, the paper is moistened and stripped off, the final fixing process being to press into the joints a fine hard cement. It is probable that the Roman mosaists adopted this or some similar method for their more elaborate work.

Doors and Windows

As the walls of houses and similar buildings are almost invariably reduced to their lowest courses or to foundations only, the discovery of a complete doorway is of rare occurrence. A good example was found at Great Winchcombe in Gloucestershire14 early in the last century. Its height was 6 ft. 2 ins., and width about 3 ft., and the casing was of stone, the sill, lintel, and jambs being of a single piece each. More often the casings were of wood, and, needless to say, these have perished. Of the actual doors perhaps the only one which survived to modern times was that of the treasure-vault at Chesters. It was of oak, bound and studded with iron, but it fell to pieces soon after its discovery. It was of strong construction, and of a kind only likely to be used in houses for the external doors.


[image ALT: zzz]
Fig. 79. — Stone Door-Sills or Thresholds, Caerwent.
(2 ft. to 1 in.)

 p270  The house doors compared favourably with ours, a width of from 4 to 6 ft., or even more being very frequent in Silchester; but as the casings there were almost invariably of wood and have disappeared, some reduction has to be made before these measurements represent the widths of the original openings. The exact sizes of the wooden thresholds are often determinable from the hollows which they occupied, and a thickness of 3 ins. and breadth of 18 ins. is not unusual in internal doors. At Caerwent, many of the sills were of stone and remain intact, and from them much can be gleaned of the construction of the casings. In Fig. 79 are shown several of these stone door-sills. In A, the surface generally has been hewn down so as to leave a rim along the front to serve as a sheath or stop for the door, and raised strips at the ends to carry the jamb-pieces, which were undoubtedly of stone. The two pivot-holes show that the door was of two leaves. In the remaining examples there are no pivot-holes, the doors having been hinged. Hinges such as we use for doors are seldom found on Roman sites, but 'hook-and‑eye' hinges, similar to those now used for field gates and outhouse doors, are frequent. The sills, B and C, have their ends rebated to receive the ends of the jambs. The former has a raised rim, but the latter a groove instead, to receive a wooden sheath about 2 ins. thick. E is unusual in having longitudinal sockets to receive the tongues of the jambs, which in this case must have been of wood, and at least 7 ins. across the face. D and F represent another type, having transverse grooves near the ends to receive the ends of plank jambs, and in the second of these the jambs were splayed. In three of these Caerwent sills there is a bolt-hole a trifle to the left of the centre, showing that, as in the first example, the doors were of two leaves, and this appears to have been a general custom.

[image ALT: zzz]
Fig. 80. — Stone Threshold, Housesteads.
(5 ft. to 1 in.)

 p271  Of the windows we know less than the doors. Three imperfect examples, two of them small, were noticed on pages 205, 211-2, and it will be observed that they occurred in baths. Not a single example of a domestic window is left to us; and this is noteworthy, for the walls of houses have occasionally been found high enough to show the window-sills, had these been as near the floor as in modern houses. The lowness, as also the large size, of the modern windows is the result of cheap, clear and colourless glass; they are now as much to see through as to admit light. But as recently as a century ago, they were habitually placed higher than at present, and were small; and the medieval windows were still higher and smaller, glass being not only costly and irregular and somewhat obscure. Nothing was to be gained by having their sills low; on the contrary, high windows were better calculated to give an evenly diffused light, and their sole function was to admit light. In the Pompeian houses the windows were, as a rule, few, small, and placed high; but before the destruction, larger and lower windows had come into vogue.

The window-glass found on our Roman sites is rarely less than ⅛ in. in thickness and of a greenish-blue tinge. That it was cast in rectangular slabs is proved by the rounded edges often seen on the fragments. The face produced by contact with the bed on which the molten 'metal' was poured, is flat and dull, while the other face is wavy and bright; and the former face was probably rendered more obscure by scouring with sand. These slabs were cast in the requisite sizes for use.15 Another variety of window-glass is thinner, equally bright on both faces, and somewhat streaky, and it resembles the glass still to be see in old cottage windows. There is some evidence that it was later than the cast glass.

It is supposed that iron cross-shaped objects, which have been found at Silchester, Caerwent and elsewhere, were used for fastening glass in windows. They are perforated in the centre and are from 7 to 8 ins. across. They were probably fixed at the intersections of the bars by nails, rivets or screws, as indicated  p272 in Fig. 81, the arms being diagonal Zy, and placed so as to clip the corners of the panes.

[image ALT: zzz]
Fig. 81. — Cruciform Iron Object (¼), and its probable use for holding window-glass in its frame.

Heating of Rooms

There were three methods of heating rooms and Roman Britain — by hypocausts, by fireplaces, and by braziers. The last was the most ancient method, and was in common use in Pompeii for heating the living-rooms of houses, the hypocaust being confined to baths. But in Britain, the colder climate brought into general requisition the hypocaust, and thus limited the use of the brazier. Internal fireplaces, which may be regarded as fixed braziers, appear to have been exceptional.

The principle of the hypocaust is so well known, and has been so frequently referred to in these pages, that it is almost superfluous to say that it consisted in the utilization of the hot gaseous products of an external fire by conveying them through flues or other spaces under the room or rooms to be warmed. This implies some means of drawing these gases through these spaces. This could have been effected by a vertical chimney at the end of the hypocaust furthest from the fire, but the Romans improved upon this by having a number of small chimneys or flues distributed  p273 along the walls of the apartments, so that not only the floors, but the walls, radiated heat. It is evident that by this means a large percentage of the heat of the fire was utilized, contrasting in this respect with our open English grates, from which a large portion escapes by the chimneys. The English system, however, ensures ventilation, which the Roman did not; but as the Roman rooms were large and probably lofty, this defect may not have been seriously felt; besides, there may have been means of ventilation we know nothing of. The cheerful glow of the English fire was also lacking, but in compensation there were the gay colouring of floors and walls and an equable temperature. Compared with our stoves and hot-water and low-pressure steam pipes, the hypocaust had a decided advantage, as the gentle heat of floors and walls could not have communicated to the air the scorched odour so often experienced in rooms heated by this means. The chief defect of the system was the absence of means of regulating the heat, except by raising or lowering the fire of the furnace; and from the great and necessarily thickness of the floors and the low conductivity of their material the effect of this could only have been felt after many hours.

In its simplest form the hypocaust was a drain-like horizontal flue under the floor of the room or set of rooms to be warmed, the mouth of which was in one of the external walls, and formed the stoke-hole or furnace, and the opposite end branched and terminate in several wall-flues. Simple hypocausts of this kind were constructed even in the largest houses; but more frequently the distribution of the heat was improved by a regular system of branch-flues. One common device was to carry the main flue to the centre of the apartment, from which point branches radiated, Union-Jack fashion, to the corners and sides of the room, and communicated directly, or indirectly through the intervention of a surrounding flue, with the wall-flues.

Another form, which is perhaps more frequent, was a shallow basement the size of the room, varying from 2 ft. to 2 ft. 6 ins. or more in height, into which the furnace opened, and from the sides of which the wall-flues arose, the floor of the apartment being supported upon a multitude of pillars (pilae). The advantage of this form was that it allowed of a more equable diffusion of heat, by exposing a maximum and evenly distributed area of the under surface of the floor to the heat; hence it was used for  p274 rooms which required a considerable degree of heat, and especially bathrooms. Although less simple and demanding greater constructive skill, it was the earliest form of hypocaust, and is the one described by Vitruvius and other classical writers.

We have thus two types of hypocausts in Britain — the  p275 'channelled' and the 'pillared.' The two types may be combined, thus giving rise to the 'composite' hypocaust, the main flue of which opened into a small pillared space, from which other flues radiated to the walls of the apartment. In Fig. 82 are shown the plans of a radiating channelled hypocaust and a composite one, both Silchester examples; while in Fig. 83 is presented both a view and a section of the lower part of a room with a pillared hypocaust below. The furnace or stoke-hole is seen on the right, and the furnace-house beyond, which would be descended by steps.16 In the opposite or left wall of the room is shown a wall-flue in section.

[image ALT: zzz]
Fig. 82. — Plans of Channelled and Composite Hypocausts at Silchester.
(10 ft. to 1 in.)

[image ALT: zzz]

The flues of the channelled hypocausts were usually from 1 ft. to 1 ft. 6 ins. wide and as high or higher, with sides built of rough masonry set in mortar or clay, and spanned above with flagstones or lag tiles, upon which the concrete of the floor was laid. The floors of the flues appear to have been, as a rule, the natural soil hardened by beating. In the pillared hypocausts they were usually of concrete, so as to provide a firm foundation for the pilae. These pillars were constructed of tiles, blocks of stone, and, very rarely, of long box-flues set on end. In districts where suitable stone was not obtainable, as at Silchester, tiles were used, and even where such stone was plentiful, there was a decided preference for these, as they withstood the heat better than stone. The tiles were generally about 7 ins. square, laid in mortar or clay, and the stack rested upon and was capped with a single one of larger size. In a few instances octagonal tiles were used, as in a Silchester example; and it is not unlikely that the pilae were sometimes cylindrical, as round and half-round tiles have been found on Roman sites. When stone was used, the support usually consisted of a single block roughly squared; but sometimes they were more carefully wrought with a projecting plinth and cap. As these hypocausts were roofed with flagstones or large tiles, the pillars were necessarily near one another, being rarely more than 18 ins. apart.

The furnaces were often constructed of tiles, but if of stone, the sides and top were coated with brick-mortar to protect them from the action of fire; and for the same reason the sides of stone flues and pilae, where exposed to great heat, were similarly treated. The furnaces of baths projected several feet in order  p276 to provide supports for the tanks or boilers which supplied hot water for bathing purposes.

The wall-flues were constructed of special tiles for the purpose, and these were of two forms, the one like a box open at the ends, and the other like a roofing-tile with the flanges more produced (Figs. 84 and 85). The 'box' tiles were generally oblong in section and about 1 in. in thickness, but varying in dimensions; a length, however, of from 1 ft. to 16 or 17 ins., and a width of from 6 to 8 ins. across the face, and of 4½ or 5 ins. across the sides, were usual. The faces were almost invariably scored with a point or a 'scratch,' so that the mortar which held them to the wall and the stucco-facing of the apartment might firmly adhere. The scoring took the form of some simple pattern such as the whims of the makers suggested. Very rarely the necessary unevenness was produced by a stamped device of an ornamental character. A flue-tile thus ornamented with an elaborate floral design may be seen in the Guildhall Museum. These tiles were moulded by hand round blocks of wood, which, upon completion, were withdrawn; less frequently, in reverse manner, within wooden moulds.

The wall-flues were, as already stated, usually placed at intervals along the sides of the apartments, in which case they were let into grooves and were completely hidden by the stucco. When it was necessary to line a wall with flues, as in the hottest rooms of baths, they were simply built against the wall, to which they adhered by mortar; but often the precaution was taken further to secure them by T‑shaped holdfasts. In such a jacketing it was usual for the tiles to have lateral openings, oblong or circular, in order that the hot air might freely pass from flue to flue, and so materially help to equalize the temperature. In Fig. 84 are shown six examples of box-tiles. The fourth was at the bottom of a stack, and has the lower portion of its front chipped out into an arched form, so as to enlarge its opening into the hypocaust. The third is a socketed variety from Gellygaer, and is perhaps unique. The flanged flue-tiles were placed with their edges next the wall, and when combined to form a jacketing were probably always held in position with holdfasts. The lateral openings were provided by cutting or scalloping out portions of the flanges, as shown in Fig. 85, where also are shown two holdfasts.

[image ALT: zzz]
Fig. 84. — Examples of Box Flue-Tiles. The third is a rare form from Gellygaer. (all ⅛.)

[image ALT: zzz]
Fig. 85. — Flue-Tiles, Llantwit Major and Caerwent. Also T‑headed nails for securing flue-tiles to walls. (All ⅛.)

 p277  Internal fireplaces may have been more numerous than is supposed. Several have been found at Silchester, notably in House 2, XIV, one of which had a tiled hearth about 2 ft. square, with cheeks and apparently a back recessed in the wall, formed of upright tiles. Others have been found at Bignor, Colerne, and elsewhere. Nothing is known of the chimneys of these  p278 fireplaces.17 Central hearths were not uncommon, and examples in the headquarters of the forts were noticed in Chapter IV.

To what extent braziers were used in Roman Britain is uncertain. The writer cannot recall having seen in any collection of our Roman remains a brazier, or anything that can reasonably be construed as one, and if, as probable, they were made of sheet-iron, their disappearance can be accounted for. But there is indirect evidence for their use. It is noticeable that in the remains of the houses of the period hypocausts are confined to very few rooms — apart from bathrooms — and occasionally are quite absent; also, that in many instances they did not form part of the original structures. As fireplaces are of rare occurrence, the conclusion seems to be that braziers were much relied upon for heating purposes. More definite is the evidence of burnt patches on the floors of rooms. In the centre of a mosaic pavement in House 1, XIV, at Silchester,18 was observed such a discoloured patch, the central portion of which, nearly circular and about 18 ins. in diameter, had been carelessly relaid with new tesserae. A portion of a pavement from Caerwent, now in the Welsh Museum, shows a similar relaid patch with indications  p279 of the effect of fire around it. It is a reasonable inference that braziers were habitually placed upon these spots. Cognate with these, are the central hearths observed in some of the headquarters' rooms of the forts (p84); in these cases we may reasonably infer that charcoal fires were directly raised upon them.


The great aqueducts that supplied Rome, Nîmes and many other continental Roman towns with an abundance of pure water, were among the most remarkable feats of ancient engineering; but there are no remains of the kind in this country that will bear comparison with them. There is evidence that some of the northern forts were supplied with water conveyed from a distance, as noticed on pp106‑7. Too little is known of our Roman towns, with the exceptions of Silchester and Caerwent, to throw any light upon their water-supplies. At Silchester, underground pipes have been found, but only sparingly so, and there is no evidence that they distributed water from a source outside this walls. On the other hand, wells are so numerous that there is little doubt that the Callevans relied upon them almost entirely for water. At Caerwent, pipes are more numerous, and wells less so, than at Silchester, rendering it probable that some of the supply was external. Wells have been frequently found within or in the near vicinity of the rural houses; and there is evidence that these houses were occasionally supplied from springs by pipes or conduits.

Many of the Roman wells with steined or lined with stone. Some of the Caerwent wells were thus lined from top to bottom; others only part way down, the gravelly clay below being sufficiently hard to require no support. At Silchester the normal soil is lighter, and while it was occasionally solid enough to require no support, there was generally a lining of flints. Almost invariably, however, the well-sinkers encountered a seam of variable thickness of wet sand which had little cohesiveness. To retain this in place, they usually resorted to timber framing, which consisted of planks of scrub oak or alder about 3 ft. long, laid in courses of four arranged rectangularly and crossing at the angles as in an 'Oxford' frame. At the points  p280 of intersection they were notched into one another and so held in place, and clay was pugged between them and the sand. Less frequently a large barrel with the ends removed served the same purpose, and in several wells two barrels were found, the one above the other.19 One well was steined with wattling, held in place by a circle of stakes driven into the clay bottom.20 Wells lined with timber-framing as above appear to have been found near Covehithe and Felixstowe in Suffolk, and at Ashhill, Norfolk.21 Medieval barrel-lined wells have been found in London.22

The pipes for the distribution of water were usually of wood or lead. The iron ring-joints of tree-pipes have been found at Silchester and Caerwent, but of the pipes themselves nothing remained. The positions of the joints indicated that they varied from 5 to 7 ft. or more in length. The joints from both places are precisely alike and vary from 3 to 4½ ins. in diameter. They are hoops with sharp edges and a mid-rib or stop, as shown in Fig. 86. The sharp edges allowed them to be driven into the ends of the tree-pipes, the stop insuring their equal penetration. The lead pipes were made of strips of sheet-lead curled into a cylindrical form until the edges met; but the method of uniting the edges varied. The simpler method was to solder them together, and a piece of pipe, 3 ins. in diameter, thus made was  p281 found at Gellygaer. But the usual method was to turn back the edges of the strip sharply before bringing them together, and to unite them by 'burning' with lead; that is, molten lead heated to beyond the melting-point was poured along the joint, when the superheated metal fusing the edges of the joint became part and parcel of the structure of the pipe. Such pipes are distinguished by the ridge which marks the joint.23 Socketed water or drain pipes of coarse earthenware have been found at York, Castor, Droitwich, Lincoln and other places.

[image ALT: zzz]
Fig. 86. — Iron Joint for Tree-Pipe, and section showing its use. (⅕.)

Treatment of Internal Walls

The walls of apartments were almost invariably plastered and painted. According to Vitruvius, the plaster (tectorium) was to consist, first, of three coats of sand-mortar (arenatum), each finer than the last, followed by two thin coats of marble mortar (marmoratum, a mortar containing pulverised marble instead of sand), the second being very fine; and while this was still moist, finely ground marble was to be worked in until a solid and smooth surface was produced. When the surface was hard and dry, it was polished with chalk and lime until it shone like polished marble. In our country no plaster of this elaborate manipulation has been found. What we generally find is one coat or more of sand-mortar resting directly on the wall and worked up to a smooth face with lime. In the chief rooms the walls often had a quarter-round or angle moulding of mortar along their foot. This skirting was painted, but not always: the writer observed one at Caerwent which retained patches of a fine surfacing resembling Keen's cement, but which had not been painted like the wall above. From the statements of Vitruvius and the remains at Pompeii, it is clear that the Romans were familiar with plastered ceilings; but nothing has been found to indicate how the ceilings were treated in this country, although it has been surmised that some of the broken plaster on the sites of rooms at Silchester may have fallen from them.

The painting of the walls is remarkable for its fresh appearance, in spite of its long burial in damp earth. This is due to the fact that it is fresco. There are two kinds of fresco in vogue now  p282 fresco buono and fresco secco. In both, the colouring matter is mixed with water or some watery solution; but in the one it is applied while the plaster is still wet, and in the other, after it has dried, the union in this case being effected by soaking the stucco with lime-water before the application. In each, but especially in the first, the colour is chemically united to the lime of the mortar, in this respect differing from distemper and oil-paint. That fresco buono was well known to the Romans is proved by the statement of Vitruvius: "Colours when carefully applied on moist stucco do not therefore fade, but last for ever. Stuccoed walls when well executed do not easily become dirty, unless the painting was carelessly executed after the surface was dry." In Roach Smith's Illustrations of Roman London24 is figured a busy scene on the tombstone of a painter at Sens, in which the deceased is represented as engaged in his craft. It so well illustrates the process that we give the description verbatim: "The subject represents the decoration of a corridor in fresco painting. A low scaffold is constructed, partly on tressels, and partly resting upon the basement of the corridor. Upon this scaffold are the painter and his plasterer. The latter is on the right side of the relief, and is exhibited in the act of laying on the thin finishing coat of plaster (intonaco) for the painter, who is following him. He has his float in his left hand, while his right hand is thrust down into a pail of water, most likely to reach a brush to sprinkle the rough coat or ground, so as to render it sufficiently moist to receive the intonaco, or thin cement of lime, which in general would not be thicker than a crown-piece. The painter is following the plasterer, to lay on his colours while the plaster is still wet. He appears as if resting one foot upon a stool, which, perhaps, has also a tablet of mixed colours upon it. Behind him is a cylindrical box, in which, it may be imagined, he has his rolls of paper or parchment with designs of the work he is engaged upon. There is a short ladder to mount the scaffold, by the side of which is a stool, with a tablet of colour upon it; and close by this the painter's assistant is mixing tints; and his action is energetic, no doubt to indicate haste. This is quite in accord with the modern practice of fresco-painting, which requires every department to be conducted with rapidity as  p283 well as with skill, The assistant must always have the tints ready mixed, and in sufficient quantity for the work. Under the arch of the corridor, at the left side of the relief, is the director or master-designer. He is seated with an open book or tablet before him, and appears to be studying or reviewing the design."

Whether this was strictly the process in Roman Britain is uncertain. Messrs. Hope and Fox consider that the Silchester mural paintings are fresco secco; and certainly the distemper-like film of colouring, sharply defined, but not easily separated from the surface of the stucco, which may be frequently noticed, favours this view. Perhaps both methods were employed. In fresco secco it is customary to rub the stucco with pumice in order to make it more pervious to the lime-water, and the same end may have been gained by the scraping with a toothed tool which our Roman examples often underwent before the application of the colours. The medieval frescoists were well aware of the high solvent powers of saccharine solutions upon lime, and used milk or infusions of sweet worts for mixing their colours; and it is likely that the Romans were also acquired with this property. The effect of fresco secco is not so brilliant and transparent as that of true fresco; but it has the advantage of requiring less haste in its execution.

The colours used in fresco are necessarily such as resist the action of lime. In Roman Britain they were fewer and less costly than those of the Pompeian and other Roman frescoes of Italy. The reds consisted of oxide of iron in some form or other; terra vert was used for green; lapis lazuli for blue; yellow ochre for yellow; chalk or a fine white lime for white; and carbon in some form for black. Chalk was mixed with most of these, and its effect was to lighten the shade, and in the case of blue to produce a blue-grey; while various blendings of the colours gave rise to chocolates, browns, purples, etc. It is probable that vegetable colours were also used.

The walls were sometimes painted in monochrome, red being preferred; but polychromatic decoration, even if of a very simple character, was the rule. Messrs. Hope and Fox, in describing some of the Silchester designs, state that "after the walls had received their finishing coat of plaster, the setting-out lines of the decorations were drawn upon the surface of the wall with some sharp instrument. The ground colours were then applied,  p284 and the incised lines showing through them served as guides for the application of the ornamentation. Traces of these setting-out or guide lines are to be seen in the fragments last referred to, and, as they are filled by the ground colour, they must have been incised in the plaster surface before it received any colouring. This process appears to have been used in decorative painting in Pompeii."25 These incised lines, however, were not always used. It is probable that the outlines of the chief features of the designs were sometimes indicated by chalk or crayon; and in the case of simple rectilineal designs, the painter may have relied solely upon the straight-edge.

The designs were more varied in style, and were generally lighter and freer in treatment, than those of the mosaic pavements. This was due in part to their execution with the brush, which allowed of freedom of treatment and fineness of detail; and in part to a sense of artistic appropriateness which recognized that the decoration of walls should contrast in its light and airy effect with that of floors made to walk upon. Moreover, they differed, certainly as a rule, in kind. The usual geometrical framework, with its bands of braidwork and twist, of the floors, was rarely, if ever, reproduced on the walls. A dado was a common feature, and the decoration of this was appropriately heavier in character than that of the wall above, panelling, painted to represent an incrustation of marble, being of frequent occurrence; whereas the fragments from the wall above often exhibit stripes and foliage, and occasionally animal and human forms. A frieze next the ceiling there may also have been, but our country provides no clue as to its decorative treatment. A few examples will give the reader a better idea of the mural decorations of the Romano-British home than a general description.

In a Roman house at Ickleton,26 the prevailing wall-colours were "red; red and white, with black stripes; blue; a greyish blue spotted with red and yellow; yellow, red, and white." The walls of some of the rooms had a rich red ground divided into panels by borders of various colours, in which were interspersed birds, flowers, stars, and fanciful objects; those of other rooms were ornamented with human figures, or nymphs and genii. The spotted work referred to was of common occurrence, and was produced by splashing from a brush. At Acton  p285 Scott,27 "fragments of decorated painting showed that the ground had been white or very light-coloured, upon which panels appear to have been marked out by lines of dingy purple and red, the ornaments being round spots arranged by fours and fives pyramidically. On one fragment was painted the head of a bird with a branch in the beak, indicating that ornamental designs had been painted on some of the panels." According to Roach Smith, the Roman mural decorations in London were, as a rule, finely executed, and consisted usually of panels of red, dark grey or black, with borders of various colours, the panels often containing arabesques, with birds and other animals, stars, and fanciful objects; and at Great St. Helens a red ground with bordering of white and dark blue or purple was covered with a delicate trellis pattern in yellow, starlike flowers in yellow or white, etc.

The Silchester reports, as might be expected, throw much light upon these decorations. In one room these included "golden-coloured draperies and imitations of yellow and grey marbles, no doubt suggesting the marble wall-linings of important buildings." In another, "grounds of yellow, red and blue, with traces of lines and ornaments on them, and painted imitations of marble, apparently a breccia." Another was "adorned with paintings representing panels of different coloured marbles separated by bands of porphyry; the marbles being a pale grey, almost white, with bluish streaks or bands, and a yellow, with blotches of pink and dark red veins." In another, "brilliant red panels with purple borders seem to have covered the walls, while other fragments showed grounds of gold colour, blue, and green."28 In the corridor of another house the fragments suggested "that the general arrangement of the colouring may have been a series of panels divided by broad vertical red bands, the panels being alternately white, with painted draperies, and yellow. There may have been a dado in imitation of pink marble, and a frieze in white ornamented with green circles." The pattern on the fragments of apparently a dado "exhibited a series of rings and hollow squares of a grey colour upon a dark claret-red ground, linked together by lines of ears of barley, with intermediate  p286 centres of blue rosettes. From the circles alternating with the squares ran diagonal lines of leaves and berries, which touched at their ends the bands of green and black bordering the composition above and below. From the angles of the squares ears of barley started diagonally also."29

At Caerwent many examples of decorative wall painting have been discovered, and the large accumulation of fragments in the temporary museum promises a rich reward to careful study. In one of the houses portions of the painted dado remained, and on it "there seems to have been an attempt to represent architecture in perspective, in far-away imitation of the later Pompeian styles of decoration; but the colouring is crude in the extreme, and the intention of the artist by no means clear." This room had two, if not three, earlier coats of plaster, each decorated, and on the earliest the hands were finer and the drawing better than in the later decorations.30 In one of the large rooms on the north side of the Basilica the decorative scheme was a series of broad-panelled pilasters arising from a marbled plinth, with intervening panels bordered with yellow and on a rich red ground. The panels appear to have been in imitation of green marble, but had suffered much from decay.31 In another house the dado appears to have been of a dark blue-grey colour divided into panels, and on one fragment is shown a hare in light brown. The border above was extremely graceful, consisting of a series of interlacing arches or semicircles in red on a yellow ground, enclosing conventional foliage in green; while the walls above were in pale blue with yellow decorations apparently forming large panels on a red ground.32

The Author's Notes:

1 Arch. Camb., 1907, p196.

2 Personal observation. See also Archaeologia, LX, p443.

3 Archaeologia, LVI, p243; LVII, p234. Arch. Cantiana, XXII, p49.

4 Price, Bucklersbury, p44. See also Archaeologia, LX, p438.

5 Archaeologia, LVIII, p24.

6 Archaeologia, LX, p453.

7 Cardiff Naturalists' Society Trans. x, p49.

8 Archaeologia, LV, p420.

9 Archaeologia, LIV, p217.

10 Ib. lii, p744.

11 Ib. lvii, p230; also liv, p219.

12 Uriconium, p200.

13 Archaeologia, LV, p227.

14 Ib. xix, p178.

15 A large stone with a shallow recess, 12 by 8 ins., on one surface, found at Wilderspool, is regarded by Mr. Thomas May as a mould for window-glass. He found in the vicinity remains of Roman glass-works. Warrington's Roman Remains, p82.

16 For example, at Silchester, Archaeologia, LX, p435.

17 Archaeologia, XVIII, p213; LV, p240; LVIII, pp20, 26, 417; Arch. Jour. XIII, p328.

18 Archaeologia, LV, p226.

19 Archaeologia, LV, pp245, 414; LVI, pp107, 238, etc.

20 Ib. lvii, p94.

21 Arch. Jour. lvii, 102. Vict. Hist. Norf. i, p295.

22 Proc. Soc. Antiq. 2, xviii, p355.

23 For an account of Roman lead pipes, see Archaeologia, LVII, p411.

24 P61. See also Wright, Celt, Roman, and Saxon.

25 Archaeologia, LV, p248.

26 Brit. Arch. Assoc. iv, p360.

27 Coll. Antiq. vi, p121.

28 Archaeologia, LV, p248; LVI, p233; LVII, p241; LVIII, p19, etc.

29 Archaeologia, LV, p250.

30 Ib. lviii, pp142, 400.

31 Ib. lx, p129.

32 Ib. lx, p453.

Page updated: 16 Jul 06