Inventory of Roman London: The defences, introduction

An Inventory of the Historical Monuments in London, Volume 3, Roman London. Originally published by His Majesty's Stationery Office, London, 1928.

This free content was digitised by double rekeying. All rights reserved.


'Inventory of Roman London: The defences, introduction', An Inventory of the Historical Monuments in London, Volume 3, Roman London, (London, 1928), pp. 69-82. British History Online [accessed 21 June 2024].

. "Inventory of Roman London: The defences, introduction", in An Inventory of the Historical Monuments in London, Volume 3, Roman London, (London, 1928) 69-82. British History Online, accessed June 21, 2024,

. "Inventory of Roman London: The defences, introduction", An Inventory of the Historical Monuments in London, Volume 3, Roman London, (London, 1928). 69-82. British History Online. Web. 21 June 2024,

In this section


(A) The Defences.

The Roman town-wall was laid out in such a manner as to enclose the two hills of London, and at the same time to take advantage of the natural contours of the ground (see above, p. 14). From the Tower northwards and westwards as far as Ludgate it remained the nucleus of the City's defences throughout the Middle Ages, but between Ludgate and the Thames it was destroyed in 1282 in order to enable the Dominicans to establish their convent across its line. The salient shown here on the old maps dates therefore from that period, and the original course of the wall can only be recovered by excavation. The evidence recorded from this source has been disputed. It will be seen below (p. 92), however, that it indicates with fair certainty a course proceeding directly southwards from the gate to the Thames. The alternative view, that the wall turned south-eastwards from Ludgate and struck the river-wall in Upper Thames Street at the foot of Lambeth Hill, is without substantial foundation. (fn. 1)

The line of the river-wall is in part less certain. Stow notes that Fitzstephen, as early as the 12th century, could only say that "on the Southside also the Citie was walled and towred, but the fishfull river of Thames with his ebbing and flowing, hath long since subverted them." Whether indeed a river-wall formed a part of the original scheme of defence is uncertain, for the fragments hitherto observed— particularly the long stretch recorded by Roach Smith in Upper Thames Street— differ in structure from the Roman land-wall and may safely be ascribed to a later date (see below). For a considerable distance W. of London Bridge no town-wall has been identified, although recent excavations there have been specially watched in this connection. To the E. of the Bridge, the wall has been recognized near the Monument and possibly in front of the Coal Exchange, and there is much probability in the conjecture that the inner curtain of the Tower of London from the Bell to the Lanthorn Tower represents its eastern portion [see A. W. Clapham in Clapham and Godfrey, Some Famous Buildings and their Story, 32].

The landward-wall was fronted by at least one ditch, and may have been backed by a bank. At two later dates both it and the river-wall were armed with projecting towers or bastions, of which 17 are known from structural remains. To these may be added four others recorded on the maps of Agas, Ogilby and Morgan, etc., whilst Mr. Clapham, noting that the surviving fragment of Roman town-wall at the Wardrobe Tower in the Tower of London points straight towards the Lanthorn Tower, has suggested that this may stand on the site of the corner-bastion of the Roman defences; with the further inference that the Wakefield, Bell and Middle Towers, extending westwards at almost equal intervals, may well represent former Roman bastions along the river-wall. If this be so, the total number of Roman bastions known or inferred is 25.

Fig. 9

Before describing the recorded remains in detail, it will be convenient to summarize and discuss the various features of the Roman defensive system seriatim, with special reference to the problem of their chronology.

Fig. 10. Sections showing foundations of City-wall. From the Victoria County History of London, I, by permission.

(i) The landward-wall, with its bank (?) and ditch. The materials of the town-wall are uniform throughout its entire length on the landward side. The main structure (Fig. 9) is built of Kentish rag-stone, roughly squared and coursed on the external and internal faces, but laid in its rough state in the core and occasionally set herringbone fashion. Occasional flints also occur in the core, but no chalk or septaria has been observed. The chamfered external plinth, which is also a constant feature, is of red ferruginous sandstone, apparently also of Kentish origin, employed in varying lengths from 1 ft. to 2½ ft., 9 in. high, and 1 ft. to l½ ft. deep. Lining with the plinth on the internal face is a triple facing-course of brick, generally, but not always, with an offset between the top and the second brick. This facing-course together with the bonding courses which are carried completely through the wall at higher levels, is composed of the ordinary Roman bricks about 17½ in. by 12 in. by 1½ in. or 2 in., generally red in colour, but occasionally yellow. These materials are set in a white mortar of extreme hardness which was perhaps run into the core in a fluid state as the interstices are not completely filled. The face at the N.W. angle of the city, by Newgate, was found to have been pointed in pink mortar, and the same feature is said to have been observed at the back of the bastion on Tower Hill. The wall stands upon a foundation (Fig. 10) of flints and puddled clay, laid in a trench, 3 to 4 ft. deep, cut in the original surface of the ground; occasionally fragments of rag-stone are used in the place of flint.

So much of the structure of the wall has now been examined that it may be stated with confidence that in its original state it contained no re-used material of any sort, the rag-stone, sandstone and brick having all the appearance of new material accumulated for the express purpose to which they were put.

That the wall was built in lengths, according to the usual Roman fashion, is indicated by the fact that there are numerous breaks in the levels of the bonding courses which are themselves not of uniform thickness or spacing; such breaks occurred between sections 9 and 13, 13 and 20, 22 and 26, 31 and 33, 33 and 35, and no doubt elsewhere, but there is no record of the observation of any one of the junctions between them, which, no doubt, sometimes occurred at the gates.

The thickness of the wall above the plinth varies between 7 and 9 ft., the thinnest sector being that by the Wardrobe Tower (1) and the thickest near Castle (Goring) Street (15). Normally the wall is reduced in thickness at each bonding-course by an internal set-back of some 3 in. At one point, however, under the Southend Railway (7), these set-backs occurred on both faces of the wall, as is attested by three independent observers. The reason for this variation from the normal, which can only have continued for a short distance, is not apparent, but the wall at this point passed over a channel or water-course of some description.

The original height of the wall is not now ascertainable, the highest definitely recorded piece (16) standing 14½ ft. above the plinth and containing one triple and three double bonding-courses. This would appear, however, to have been exceeded by the portion in Trinity Place (as recorded in Fairholt's engraving) which was perhaps a foot or two higher.

Whether the town-wall was originally backed by an earthen bank is uncertain. The better preservation of the internal face has been remarked more than once, but may well be ascribed to the protection afforded by the inevitable accumulation of soil and débris within the walls of an ancient city. On the other hand, the actual traces of a constructional bank were suspected in one of the cuttings on the site of Christ's Hospital (below p. 91, Fig. 16), and the presence of banks in conjunction with the thick stone walls of Colchester, Caerwent, Silchester, and other Romano-British sites strengthens the suspicion. The fact that such a bank has not otherwise been detected in London is of little weight, since not only must a bank in this position have suffered drastically from post-Roman builders, but, even when it has survived to modern times, none but a specially trained observer could be expected to detect it.

Traces of a single V-shaped Roman ditch, cut in the gravel, have been observed at two points to the E. and two points to the W. of the Walbrook. This ditch lay at a distance of 10½ to 15 ft. from the base of the city-wall, was 10 to 16ft. wide and 4½ to 6½ ft. deep. Any outer ditch or ditches which may have existed were presumably removed in the cutting of the large mediæval ditch, which had indeed at two points almost obliterated the Roman ditch above mentioned. At Aldersgate Street a ditch 74½ ft. wide and 14 ft. deep, at one place with a broad flat bottom and at another with a slight median mound (thought to be for a bridge) has been regarded as Roman. It may represent a local re-cutting at the time when the bastions were added here, but it is perhaps more probable that the Aldersgate ditch was post-Roman.

The question of the date of the landward wall of London has given rise to much discussion, often of a highly theoretical character. It is not necessary to consider the various views in detail, but in order to indicate the present position in the matter three representative theories are here summarized.

One of the contributors to the Victoria County History argues that the construction of the Ermine Street was subsequent to the first decade of the 3rd century, since it is not included in the Antonine Itinerary, which was possibly the work of Caracalla. But the Ermine Street now bends slightly westwards as it approaches Bishopsgate, and therefore "if Bishopsgate was a Roman entrance into the city, the London portion of the Ermine Street must have been shifted a little to the west when the City was fortified, and the date of the Wall is thus given within certain limits. As coffins both of lead and stone are only found beyond the fortifications, and, on numismatic and epigraphic grounds, are referable as a class to the 4th century, the Wall would seem to have been erected in the Constantine period. . . . It is to this conclusion also that the comparison with walled towns on the Continent and an examination of the structure itself inevitably lead, and there can be no better reason assigned for the bestowal of the name Augusta than this transformation of a trading town into a fortress" [V.C.H. London, I, 37].

Another contributor to the same work holds a different view: "Many of the remains . . . show by their position on the original surface, and by the coins and objects associated with them, that London had extended over a large area at a very early period. In this way the City had spread itself out, probably as an open town, its streets not disposed in any regular plan, and its buildings interspersed among trees and streams. . . . After a period, it was found necessary to provide the rapidly growing town with more adequate defences, and the wall was then built. Its irregular course indicates its adaptation to a state of things already existing, but it was evidently carried well beyond the more densely inhabited parts, as everywhere it has been found to rest on the natural surface." Then the soil of the City began to rise, and the passages prepared to carry the Walbrook through the wall became choked. "Striking evidence of the great rise in the soil after the building of the City wall, but in Roman times, is afforded by the gate at Newgate," which rests on made soil some 4½ ft. above the base of the adjacent wall. "The projection of the gate from the wall also betokens its late age, as does the employment of oolite for the plinth. . . . All these considerations lead to the conclusion that the City wall was built at a far earlier period than has generally been conjectured, and an approximation to its date may perhaps be furnished by the coins found in the bed of the Walbrook. Many of these have been found, which all fell into one group ending with Marcus Aurelius and the Faustinae, and may fairly be taken to indicate the period when its bed and the vicinity ceased to be occupied after the choking up of the stream. From this it would appear that the wall was built at the latest by the middle of the 2nd century" [Op. cit. 77–9].

Lastly, Professor Haverfield's summary of the problem may be cited: "The material facts appear to be the following. First, the wall itself, as I have stated, stood entirely on the clean gravel subsoil; no sign has been noted that it anywhere crosses earlier buildings or even graves. This would suit an early date, though perhaps our records are too imperfect to give us certainty, at least in the matter of such small remains as burials. Secondly, the line of the wall was seemingly laid out so as to include almost all the buildings of the town which were in existence when it was constructed. The area thus enclosed is very large, and this suggests considerable growth and a late date. Thirdly, the absence of buildings, other than graves, close outside the wall suggests that it was built late in the life of the town; there was no subsequent period during which more suburbs could develop. Fourthly, the structure of the wall, with its courses of small stones and its bonding-tiles, is generally ascribed to the later empire. This, however, is not quite so certain as is generally alleged. Bonding-tiles were certainly used very freely in the later empire. We have datable instances in the forts of the Saxon Shore, and in the city of Trier and on other Gaulish sites. But, at any rate south of the Alps, they were known and used in the 1st century A.D. They may favour, but they certainly do not prove, a late date for the London wall. Lastly, we have the general probability that wall-building would be carried out in Britain, as in Gaul and elsewhere in the empire, under the pressure of some evil, such as the attack of the barbarians. It is plain that these considerations permit of no definite conclusion. We must wait and see. But some of the facts appear to have rather more weight than others. In particular, the large area included by the wall, and the scarcity of dwellings outside it, and the need of some historical cause for wall-building, combine to make me think that perhaps the end rather than the beginning of the 3rd century is the more probable date. The bastions might easily have been added in the course of the 4th century, when the dangers from Saxon pirates became even more acute " [Journ. Roman Studies, I, 158].

Many of the premises or inferences which have been quoted above are open to dispute, and it is now desirable to review the whole of the evidence independently of them. It may be approached from three different aspects:—

(a) The actual evidence provided by the remains.

(b) The general evidence provided by the form and compass of the wall, with special reference to burials within and without its circuit.

(c) The general evidence of analogous sites in Britain and Gaul.

(a) No fragment of re-used material has been recorded from any part of the landward-wall. One observer (Dr. Philip Norman), who has seen at various times something like 400 yards of the wall and has particularly looked for such evidence, declares that he has seen none, and this testimony combined with that of all other records enables us to assume that the structure was built throughout of new materials.

The character of the construction, particularly the use of the lacing-courses, suggests comparison with late Roman fort-walls, such as those at Richborough, Lymne and Bradwell. On the other hand, this construction occurs freely in early walls, e.g., at Pompeii and Herculaneum before A.D. 79, and at Colchester, the walls of which seem to belong to the latter part of the 1st century [Trans. Essex Arch. Soc. (N.S.), XV, 179; Journ. Roman Studies, IX, 144; and cf. Haverfield, as quoted above]. The use of bonding or levelling courses would obviously be convenient in a district such as London, where freestone is not available and the material is naturally rough and irregular. It is not, therefore, a feature of chronological value.

The wall is built uniformly on the original surface of the ground, and in no recorded instance disturbed a pre-existing deposit or building. (fn. 2) This in itself, of course, only proves that, whenever it was built, it was erected outside the inhabited region (fn. 3), presumably with a view to expansion within its limits rather than with a view to the defence of the most vulnerable area.

More striking is the accumulation of 6 to 7 ft. of deposit, containing only Roman remains, against the wall and over the ditch in America Square. That this accumulation was deposited in the Roman period and after the building of the wall seems certain, but the length of time implied thereby is quite uncertain, for, while normal conditions might imply the passage of centuries, there is no evidence that normal conditions obtained in this particular case.

The rise in the Roman ground-level after the building of the wall is also indicated at Newgate, where the plinth of the Roman gateway, as found, was 4½ ft. above that of the adjacent town-wall, and evidently represents a Roman rebuilding.

The bastions, which are all or mostly Roman, are also additions, since they are not bonded into the wall and either oversail or cut through the Roman ditch; but here again there is little or nothing to show the lapse of time between the building of the wall and that of the bastions.

The most significant piece of evidence is provided by the silting-up of the Walbrook. As noted above (p. 16) the deposits in or close to the stream-bed have been observed by competent archæologists at two adjacent sites, one within and one without the line of the N. wall. On the S. side the lowest deposit consisted entirely of black occupation-earth, whereas on the N. side the lowest deposit consisted entirely of clean sand and silt. The boundary between the two types of deposit was apparently the city-wall, and the building of the wall would seem to provide the only explanation for the essential difference in the nature of the two deposits. The earliest deposit of black occupation-earth, on the S. side of the wall, is dated by its contents to the first century of the Roman occupation. A shaft sunk by the Society of Antiquaries against the outer face of the Wall, was undertaken with the express purpose of ascertaining the conditions of the deposit in the Walbrook bed at this point. The section obtained proved that the 4 ft. of sand and silt lying above the natural gravel had been deposited against the face of the Wall after its erection, there being no trace of a cutting through this deposit for the building of the Wall. This point is of capital importance, as the observations of Mr. F. W. Reader (see p. 147) made on the site 150 ft. away on the same side of the Walbrook channel, enable us to date the lowest deposit of silt with a considerable degree of certainty and to state that the lowest 18 inches of it were deposited before the construction of the pile-structures, c. A.D. 100–130. It is obvious, therefore, that the Wall must have been built an uncertain time before the pilestructures, the duration of which is represented by the deposit of 18 in. of silt which underlay them. Any other conclusion can be based only on the invalidity of the recorded evidence of the Antiquaries' shaft and Mr. Reader's observations.

(b) The fact that the wall, whenever built, amply enclosed the then occupied area has emerged above (p. 75 n.). It is a fair assumption from Tacitus that London, like Colchester, was without defences at the time of the Boudiccan revolt. (fn. 4) But by that time, as we have seen in a previous Section (p. 31), the eastern and part of the western hills were already occupied. It was, therefore, clearly necessary that any comprehensive scheme of defences subsequent to that date should enclose the two hills. This the wall in effect does, and its main lines were suggested by the natural contours of the ground (see above, p. 14). The waste ground implied by the Walbrook valley, together with the re-entrant at Smithfield, thus helped to stretch the London wall to a length which was at the time clearly greater than was required by the actual size of the town. Other factors which may have contributed to this result will be discussed below.

The relation of the line of the wall to the Roman cemeteries is of obvious importance in view of the Roman law (not always closely observed) that the dead should be buried without the precincts of a town. With the exception of a group of burnt burials, of which no details are known, found at the beginning of the 18th century just within Bishopsgate, and a large and almost exclusively early (pre-Flavian) cemetery in the area N. and N.W. of St. Paul's Cathedral, all the cemeteries of Roman London appear to have lain just outside the line of the walls. The distribution of inhumation-burials is especially significant. The rite of inhumation began to come into vogue in the latter half of the 2nd century (see above, p. 30), and only four scattered inhumation-burials which can with any probability be regarded as Roman (and one or two of these may be Saxon) have been recorded within the walled area. Equally striking is the manner in which, at the Minories on the E. and at Newgate on the W., cemeteries used extensively from the end of the 1st century onwards crowd up to the town-wall as though that were their natural limit. The evidence in the N. at Moorfields and Bishopsgate is less clear, because the date of the cemetery found within Bishopsgate is quite unknown, and the exact location (whether within or without the wall) of several burials near by is uncertain. But here too the greater portion of a cemetery used from the 1st century onwards lay outside the wall. The general implication of the cemeteries as a whole is that the line of the wall was their inner boundary by the 2nd century (see also above, p. 31), but the evidence is, in detail, too vague to justify any very definite conclusions.

(c) The closest British analogies to the original wall of London are the Roman walls of Verulam and Colchester. The date of the former is unknown; the latter, which, though differing in detail, bears a close general resemblance in character and dimensions, has been noted as probably of late 1st-century date. Haverfield suggested that the diamond broaching on the wall of Aldborough in Yorkshire might indicate a 2nd-century date for the structure (fn. 5); and an extensive examination of the S. wall of Caerwent in Monmouthshire has merely shown that it is not likely to have been erected before the Antonine period or after the reign of Constantine the Great.

In our present ignorance of our own antiquities it is necessary to look to France for further information. There, fortunately, the evidence is fairly abundant, and in outline it is as follows. The town-walls built at the time of the development of Gaul in the Augustan period are of ambitious extent, built throughout of new and carefully prepared material, and designed perhaps with a view as much to future expansion as to present need. Of this kind are the walls of Nîmes (built in 16 B.C.), Fréjus, Arles, Autun, Avenches—all with an original perimeter of 4,000 to 6,000 metres. In the case of Autun (and possibly in other cases also) ambition seems to have overstepped its mark, for there is reason to suppose that the town never grew to the full limits of the enclosed area. In sharp constrast to these expansive early fortifications are those of the later Empire, built mostly, as M. Blanchet thinks, after the great German invasion of A.D. 276. The late walls are of small extent, rarely with a perimeter of over 1,500 metres, and they are invariably built largely of material re-used from earlier buildings. The general period of this series is indicated by building-inscriptions of Diocletian and Maximian (c. A.D. 300) at Grenoble; by the presence of a coin of Tetricus (268–273) in the masonry of a tower at Auxerre; by a well-preserved coin of Claudius Gothicus (268–270) in the masonry of the wall of Bordeaux; and so forth. The period is roughly that of our Saxon-Shore forts such as Richborough, the construction of which (again including re-used material) is closely similar. These late walls enclose only a small part of the inhabited area of the town —either the part which contained the principal buildings or (usually the same thing) the part which was most easily defended. The restricted circuit of these walls was due in some degree presumably to the military emergency which occasioned their construction—to act as a refuge for the population in case of attack and to be defensible by the diminished man-power available; but it also owed something to the political and social conditions of the time. Under the early empire a corporate sense of civic responsibility had made the city in every sense co-extensive with its population, and though scattered buildings would naturally be found outside the defences, the walling of the city had implied the walling of the greater part of the populated area. Under the middle and later empire, on the other hand, this corporate civic sense had gradually given place to an increasingly bureaucratic and exacting administration with, as its counterpart, an increasingly impoverished and servile population. The city was no longer really co-extensive with its citizens; it would not be a serious exaggeration to say that it was now co-extensive with little more than its executive core. The walling of the city was now, therefore, primarily the walling of its administrative nucleus; and into the new walls were built the remains of outlying temples and other structures, now left desolate by changing faith, by hostile raid, or merely by the decay of that public spirit to which in earlier centuries they had owed their existence.

Whatever the underlying motive, however, the evidence of the Gallo-Roman cities is consistent, and it enables us to propound the generalisation that in Gaul the Roman town-walls built of new material and enclosing a large area are early (1st centuries b.c.—A.D.), whereas walls built of re-used material enclose a relatively small area and are late (3rd century or later). There seems to be no exception to this rule [see A. Blanchet, Les enceintes romaines de la Gaule, 304, etc.; and Journ. Roman Studies, XVI, 191]. (fn. 6)

This rule may not be altogether without bearing upon the corresponding problem in Britain. It is not to be assumed that the historical circumstances of the two provinces were identical in detail; nevertheless, on the one hand, it is reasonable to compare in a general way the political and military situation in Britain under the Flavian Emperors with that in Gaul under Augustus, whilst on the other hand, the environment of beleaguered Britain in the 3rd and 4th centuries certainly did not differ materially in its military aspects from that of contemporary Gaul. If it be permissible, therefore, to apply our generalisation tentatively to the London wall, it is clear that the exclusive use of new material, founded on the natural gravel of the site, together with the ample extent of the defences, enclosing rather more space than was required at the time for close building, includes it in the earlier group, with Autun, Nîmes and the rest. It is certainly against all analogy to suppose, with Haverfield, that the large area enclosed by the wall suggests a later date; it points on the contrary in the opposite direction. The later phase is approximately represented by the added bastions, built in part of re-used material. To these we shall turn in a moment. (fn. 7) It may be added that any theory which postulates a late 3rd or 4th-century date for the walls of London crowds into a single century, besides the construction of the wall itself, the addition of two different sets of bastions, the erection or re-erection of the river-wall and the rebuilding of Newgate, 4½ ft. above its first level. Moreover, the building of a wall without bastions and the addition of bastions to the same wall represent two radically different principles in defensive tactics, and we know that the change from the former to the latter system took place before the erection of the Saxon-Shore defences at the end of the 3rd century.

In summary, it may now be observed that all the evidence, direct and indirect, points to or is consistent with a relatively early date for the landward wall of London. So far as observed, it never overlies any relics of previous occupation; its material is new throughout; it is prior to two sets of Roman bastions, and to a Roman gate built after a considerable mass of material had been deposited round its base; at one point 6 to 7 ft. of Roman rubbish had found time to accumulate against the wall and over the ditch; the silt which the Walbrook began to deposit in Finsbury presumably as the result of the partial obstruction caused by the wall was covered by débris of 1st or early 2nd-century date; the principal cemeteries used from the end of the 1st century onwards seem to be bounded by the line of the wall; and the large area enclosed by it, apparently in excess of the immediate requirements of the existing town, suggests comparison with the Augustan series in Gaul—a series which, on general historical grounds, might be expected to correspond to a Flavian series in Britain. If we add the probability that the destruction of Colchester, London and Verulam by Boudicca in the year 60 awakened the authorities to the weakness of the undefended cities of south eastern Britain, and the practical certainty that Colchester at least was fortified soon after the disaster, it is unnecessary to look further for an appropriate historical setting for the walling of London. It is here inferred, therefore, that the original London wall was built during the half-century following the Boudiccan rebellion.

One further point may be noticed here. The curious planning of the landward wall on the N.W. side, with its deep re-entrant angle between Cripplegate and Newgate, has never been satisfactorily explained and no definite explanation will be here attempted. It may, however, be noted that the two walls of the right-angled salient, W. of Cripplegate, are parallel or at right angles to the fairly well authenticated road between Newgate and the Walbrook-crossing at Bucklersbury (see plan, Fig. 8 and p. 48). It is possible, therefore, that this salient represents a skirting, by the wall, of a quarter already definitely laid out before the wall was built.

(ii) The river-wall, so far as it has been observed, was built or re-built upon a foundation enclosed between two rows of piles set close together, the rows being 7 ft. apart. The wall both within this enclosure and above, where it was 8ft. thick, was composed of rag-stone with single bonding-courses of brick at regular intervals of 2 ft. The long stretch, observed by Roach Smith, contained also at the base, chalk and a course of stone set in pink mortar; the stone included many fragments of re-used architectural and sculptured stones and marble, all of Roman date.

The unstable character of the river-bank or foreshore on which this part of the town-wall was built (fn. 8) would be enough to account for any differences which occur in method of construction, as contrasted with the wall on the landward side. The employment, however, of re-used material and pink mortar shows conclusively that the two structures were of different date. The presence, on the other hand, in the river-wall of this same re-used architectural material at the base, the use of pink mortar also at the base and finally the use of brick bonding courses above, all seem to point to an equation in date between the river-wall and the eastern group of bastions where the same choice of materials and method of employing them is exemplified (see below). The date of this wall will therefore be considered in conjunction with that of the eastern group of bastions.

(iii) Structural remains of external bastions have been recorded from the line of the landward wall and others are shown on 16th and 17th-century maps. On the river-side also the city is recorded by Fitzstephen to have been "walled and towered," but these defences had collapsed before the end of the 12th century.

That the bastions were an addition to the town-wall is obvious from the fact that in every case which has been examined they are built up against the pre-existing town-wall, the plinth and facing of which is carried through behind them. In one instance, at Camomile Street (10), traces were found of an attempt to tooth or bond the bastion into the wall, but this does not seem to have been generally done. The bastions display a considerable variety of structure and materials, which indicates that they were not all added at the same time. With the exception of the Wardrobe Tower (1) all the bastions examined on the E. side of the city as far as Moorgate were solid at the base and contained Roman architectural or sculptured stones re-used as building material. These bastions number seven (2, 3, 6, 8, 9, 10 and 11), and may be held to form one group, though the most westerly of them (All Hallows Vestry (11)) is the only one that stands on a rectangular platform. Certain further details appear in several of these bastions which may have been common to the whole group, though the evidence for the rest has been destroyed. The most striking of these details is the use of bonding or surface courses of bricks in the two bastions in Duke Street (6 and 7) as attested by Woodward and Maitland, and by an 18th-century drawing of one of these bastions showing four triple courses still in position (Plate 28). A single course is also shown on the drawing of the very fragmentary Castle Street bastion (9); in this case the course occurs 8 ft. above the footing of the bastion, a height which would account for the non-survival of the feature in bastions 3, 10 and 11. Patches of pink mortar were noticed as occurring sporadically in the core of bastions 3, 8, and 11. The size of the four properly recorded bastions (3, 9, 10 and 11) varied between 19 ft. and 26 ft. in diameter and 14¾ ft. to 15½ ft. in projection.

The remainder of the bastions examined in recent years differ from the group above described in two particulars inasmuch as (a) they contain no re-used architectural or sculptured stone-work of the Roman period, and (b) with the exception of one bastion (17) on the Christ's Hospital site, they are built hollow from the base. Two of the bastions at Christ's Hospital were perhaps standing to a sufficient height to show that the brick courses of the first group were not present there, at any rate to a rather higher level; but a drawing of the bastion destroyed in 1865 in Windsor Court (14) seems to show one double course in its upper part.

As to the date or dates of the bastions, definite evidence apart from their construction has been provided in only two instances. At All Hallows Vestry (11) the soil just above the platform of the bastion did not appear to have been disturbed and yielded nothing but Roman sherds, etc. The Roman ditch had apparently remained open after the building of the bastion, save beneath the actual structure. The filling of the angle-bastion at Christ's Hospital below a level of 10 ft. from the surface yielded nothing but Roman objects. (The lower part had apparently not been disturbed by the building of the bastion, and contained Roman sherds antedating it.) It would thus appear that there is evidence of a Roman date for one example, in each group of bastions, and the question arises as to which of the two series is the earlier, for it may be assumed that they are not contemporary; the use of the débris of ruined buildings and the spoil of cemeteries argues a different political and material situation from that indicated by the use of ordinary building material. Unfortunately none of the inscriptions known from the structure of the eastern bastions can be closely dated; they are in most cases probably of the 1st or 2nd century, but a sculptured head from the Camomile Street bastion is probably not earlier than the 3rd century (see p. 46). At Chester a neighbouring cemetery was robbed of its tombstones to rebuild the wall of the fortress about the year 200; and it has been remarked above that the use of material from previous buildings is normal in the basis of the towers and walls erected in Gaul in the latter part of the 3rd and in the 4th century, and it may be noted that the general character of their superstructure is also similar to that of the eastern group of bastions. The fortress built over part of the Roman town at Richborough, c. 300, also contains re-used material. Further than that the materials do not help us.

The use of external bastions and towers of semi-circular or horseshoe form is characteristic of the great urban defences of the Augustan period (as at Nîmes, Arles, Fréjus, etc.). (fn. 9) By the time, however, that Britain was sufficiently settled for extensive town-building, i.e. by the Flavian period, the age of active urban fortification in Gaul and the neighbouring provinces had already passed, and it was the military tradition, with its earthwork ramparts and internal towers, that dominated the designing of our town-defences even when stone was actually used for them. Thus it is that in no Romano-British walled town does a projecting tower or bastion, elsewhere than at a gateway, appear to have been an original feature of defences which are likely to have been built before the 3rd century. Only in that or the succeeding century, when the era of the camp had given place to the era of the castle, did the development of fortification in stone revive the old urban tradition, and with it the use of the projecting bastion, as in our Saxon-Shore fortresses. At this period a few places such as Horncastle and Caistor in Lincolnshire and perhaps Caistor-byNorwich—all in the immediate hinterland of the Saxon Shore, and possibly used then as fortresses rather than as towns—received walls and bastions of one build and of normal Saxon-Shore character. Apart from these, in one or two cases, though rarely, projecting towers were added to existing town-walls in this country; but the only examples approximately dated are the six polygonal towers added to the S. wall of Caerwent in Monmouthshire at a period which recent excavation has shown to be not earlier than A.D. 330 (fn. 10).

From this comparative evidence it is thus possible only to affirm that neither series of bastions in London is likely to be prior to the 3rd century. Which series is the earlier, it is impossible to say. On constructional grounds it might be supposed that the free use of architectural and other second-hand material in the eastern group indicates the later date for this series, since we have not merely the late analogies already cited from Gaul, but the culminating edict of Arcadius and Honorius, issued in 396 and re-affirmed in 408, authorizing urban authorities to build or repair their fortifications with materials drawn, if necessary, from disused temples and other buildings [Blanchet, op. cit., 311; and Codex Theod., XV, I, 34, 40, etc.]. (fn. 11) On the other hand, the eastern half of the town seems to have been the more important of the two (see above, p. 31), and it was the nearer to attack from the sea. The presence of danger from this source is emphasised by the apparently contemporary building or rebuilding of the river-wall (see above, p. 80). Perhaps, therefore, on tactical grounds it is easier to imagine that the eastern series, with the river-wall, is the earlier of the two, dating possibly from the latter part of the 3rd century, when the piracy which led to the fortification of the Saxon Shore was becoming acute. This, however, is mere conjecture, and the problem of the dates of the bastions and river-wall, within the general chronological limit indicated, must await further evidence.

A word may be added about the plan of the Roman Newgate—the only London gate of which any definite indication of the Roman plan has been recovered. The fact that this gate was built of different materials and that its plinth lay at a height of 4½ ft. above that of the town-wall indicated its relatively later date, but there is otherwise no direct evidence of the period to which it belongs. The fragmentary remains suggest a double carriage-way flanked by square towers which projected from 7 to 14 ft. in front of the town-wall. A gateway of this type might be of almost any Roman date; it is certainly not (as stated by one of the writers in the Victoria County History) specifically a late type. The square projecting towers occur at Aosta as early as the Augustan period and at Cologne probably in the Flavian period, and many examples could be cited from the end of the 1st century A.D. onwards [see R. Schultze, Bonner Jahrbücher, CXVIII, 293, etc., and Wheeler, The Roman Fort near Brecon, 20]. The plan therefore does not help.


  • 1. It rests upon the discovery of a wall running diagonally across the site of No. 56 Carter Lane, and the vague statement of Roach Smith that at the foot of Lambeth Hill the river-wall "formed an angle with the Hill and Thames Street." The wall in Carter Lane was not definitely recorded or seen by any competent observer and its precise angle is entirely uncertain. The wall of a Roman building discovered near the bottom of St. Peter's Hill would in any case lie outside the suggested line, and the argument [Arch., LXIII, 306] that no Roman building has been discovered in this angle of the city, thus falls to the ground.
  • 2. Only in two instances have structural remains been recorded in immediate proximity to the wall, and in neither case is there clear evidence that the wall was subsequent to them. The first of these is the long tessellated corridor recorded in 1707 by Woodward as stopping within 3½ ft. of the wall in Camomile Street (p. 111, No. 51); the other is the flooring of large tiles which extended inwards for 10 ft. from the face of the wall on the site of Christ's Hospital (p. 91). The latter was thought to have preceded the wall and may have done so; but no trace of it was observed on the other side of the wall, and it is conceivable that the pavement—the lateral extent of which was not ascertained—may have been the floor of a tower or staircase built in conjunction with the defences.
  • 3. This point is emphasised by the long straight stretches of wall between the gates. Habitation on the outskirts of an unwalled town tends to straggle along the main approaches; thus, a subsequent enclosing wall, built from point to point to include the extreme limits of occupation, would necessarily embrace large intervening areas of thinly occupied land.
  • 4. Tacitus states that the rebels deliberately avoided fortifications and sought plunder where it was unguarded. Moreover, the withdrawal of Suetonius from London suggests that the site was unfortified [Annals, XIV, 33].
  • 5. This inference is far from certain since a somewhat similar tooling occurs on fort-walls and gates in Northumberland which may well be of 3rd-century date.
  • 6. In this connection the following table, giving the approximate areas of the two groups of Gallic towns is instructive:—
    (i) Early Imperial towns:—
    Nîmes area 790 acres
    Autun " 494 "
    Lyon " 318 " (approx.)
    Vienne " 214 " "
    Fréjus " 120 " "
    To these should perhaps be added Cologne, area 239 acres. The walls of Toulouse, enclosing an area of about 237 acres, and thus belonging to the same class, are of uncertain date, but Blanchet [op. cit. 202n] quotes the opinion of M. Joulin that they were Trajanic. The walls of Avenches enclosing an area of (very roughly) 350 acres are assigned to the 1st century by Blanchet [op. cit. 141] and to the period Vespasian—Titus by Bursian [Mittheilungen der Antiq. Gesellschaft in Zürich XVI]. Trier, area 704 acres, can scarcely be cited as an analogy since its maximum expansion probably dates from the period in which it became the Imperial capital of the West.
    (ii) Late Imperial towns:—
    Strasbourg area 49½ acres
    Sens " 47 " (approx.)
    Nantes " 39½ "
    Troyes " 39½ "
    Dijon " 27 "
    Beauvais " 25½ "
    Tours " 22¾ "
    Rennes " 22¼ "
    Le Mans " 19 " (approx.)
    Senlis " 15¾ "
    Perigueux " 13½ "
    This second list might easily be extended, but all the known examples seem to fall within the same limits of area. It is clear that London, with an area of 330 acres, belongs essentially to the first rather than to the second group. Whether to the second list should be added the contracted defences of Autun and Aries, where later walls enclosed a section only of the original area, depends upon the date assigned to the restricted enceinte. At neither Autun nor Aries has this been determined definitely, although in both cases there is a probability of a late Roman date. In none other of the large Gaulish cities—Nîmes, Lyon, Vienne, Fréjus, Cologne, Avenches, Toulouse—is there any evidence for a restricted enceinte of late Roman date. It may be presumed that the original defences proved sufficient to prevent the devastation and depopulation which overtook the open towns in the 3rd century. The analogy, therefore, between the large walled cities in Gaul and in Britain is complete without the postulation of any systematic reduction of the fortified area under the later Empire.
  • 7. It may be added that the indication of town-walls on the medallion of Constantius (p. 188) has been cited as evidence for the existence of fortifications at London by the end of the 3rd century. This evidence is in itself of doubtful value, since the representation of the city on the medal is obviously conventional.
  • 8. It had fallen before the end of the 12th century, but part of it was standing at Queenhithe in 899 [Kemble, Cod., Dipl. V, 141; W. Page, London, its Origin, 131].
  • 9. These early towers were invariably hollow. Solid towers do not seem to occur in Roman architecture until the 3rd century, and are not common until the 4th. This might seem to suggest that the solid (eastern) series of towers or bastions at London was the later of the two; but the introduction of solid towers did not involve the abandonment of hollow construction, and hollow towers of the 4th century are not uncommon, as at Caerwent.
  • 10. Mr. V. E. Nash-Williams has recently found an "Urbs Roma" coin (A.D. 330–335) beneath the floor of one of them. The coin was definitely sealed by the structure.
  • 11. The bastion destroyed in 1865 on the site of Bastion House, Windsor Court, Aldersgate (p. 104, No. 14) appears to have had no levelling courses below a height of some 30 ft. above the ground. The drawing (Plate 33) shows, however, one levelling course above this height. It is a possible inference that the upper part of the bastion was an addition or reconstruction of the period when the eastern group of bastions (with their regular levelling courses) was built—an inference which of course also implies the priority of the western group.