[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]
This webpage reproduces a chapter of
Roman Roads in Britain

by Thomas Codrington

published by the Society for Promoting Christian Knowledge
London, 1903

Text and maps are in the public domain.


[image ALT: link to next section]

 p329  Chapter X: Roman Roads from London
to Silchester and the West — continued

(10) Silchester to Speen, Bath, and South Wales
(11) Speen to Gloucester
(12) Gloucester to Redbrook, and to Caerwent
(13) Iter XIII of Antonine
(14) Gloucester to Llandovery
(15) Llandovery to Llanio and Sarn Helen
(16) Llandovery to Carmarthen
(17) Carmarthen to the West

(10) Silchester to Speen, Bath, and South Wales

A road from Silchester to Speen (Spinae) appears in Iter XIII and in Iter XIV of Antonine. It issued from the west gate, and its course is indicated by a highway with a parish and county boundary for three and a half miles, beyond which there is little or not trace to Speen, where the roads branched. Beyond Speen, following Iter XIV in the reverse direction, there is little or no trace of the road as far as Cunetio. It probably followed the line of the Bath road from Newbury towards Hungerford, and continued on by Rudge​1 to the north of Hill Barn, on the south side of the Kennet valley, about three-quarters of a mile to the east of Mildenhall, where the ridge remains on the down. Sir R. Colt Hoare, in his map of Cunetio, shows the Roman road onwards to the lane at Cockatrip Cottage, along the south of a field in which many coins and fragments of pottery, etc. have been found. He also shows "an earthwork  p330 of a square form by which the parish church was surrounded,"​2 which he supposed to be a Roman camp, and called Lower Cunetio. Traces are visible, but there is good reason to think that they are not Roman. There are undoubted evidences of a Roman settlement in the more sheltered valley, but the camp on the high ground at Folly Farm must mark the Roman station Cunetio.

There are no indications of the Roman road onwards through Marlborough. Stukeley supposed that there were remains of a camp near the river in what is now the garden of the Master of the College, but R. Gale found that he was mistaken. The first trace of the road is three miles further on, near West Overton, where a short length of the ridge is shown on the old Ordnance map on the north side of the Bath road, and it is still traceable. A little further on near East Kennet Sir R. Colt Hoare observed the causeway five feet high and 18 or 20 feet wide, and the ditches were distinct in 1884. The road passed round Silbury Hill on the south side, and although it has been ploughed up it is now sometimes traceable by the eye. In 1867, to test Mr. James Fergusson's contention that Silbury was upon the Roman road,​a some sections were cut across the road, and the trenches on each side of the road were found at a distance of 18 feet apart, and were traced round the hill to the straight portion of the road on the west of it. The road there ranges with the south side of the mound, and it can be traced in  p332 dry summers. From Silbury the course is straight for two miles and a half to a point (665′) on a spur of Calston Hill. A slight curve brings the road in a mile and a half to the 700 feet contour line on the north side of Morgan's Hill, along which it is carried on a terrace about five yards wide, cut into, and embanked upon, the slope of the hill. Where it passes round the head of a coombe at Horse Coombe Bottom Wood, the banking up is considerable. On the west of Morgan's Hill the Roman road is joined by Wansdyke, an intrenchment which crosses the Wiltshire Downs for many miles, and is plainly traceable from the borders of Hampshire to this point.​b It follows a devious course, and varies a good deal in size; over Morgan's Hill it is unusually crooked, and consists there of a ditch with a rampart on the south side, and a slight counterscarp on the north. The crest of the rampart is nine or ten feet above the down, and 18 to 20 feet above the bottom of the ditch as they now exist, and rampart and ditch together measure eighty to ninety feet across. Coming over the hills it joins the Roman road at an acute angle, and Stukeley thought that there was "incontestable proof that it was in being before Roman times," and that the Roman road followed the course of Wansdyke.​3 Local antiquaries have generally insisted on the pre-historic, or at least pre-Roman age of the great intrenchment, and that the Roman road took the line  p333 of it. If so, as Sir R. Colt Hoare pointed out,​4 we must suppose that Wansdyke, which winds about through all the rest of its course, takes an absolutely straight line for three miles to Wans, and then again for 10 miles on to Ashley Wood, making together 13 miles of straight course along which a Roman road follows it. The excavations made by General Pitt Rivers in Wansdyke close by prove from coins that the work is Roman or post-Roman in date, and the earthworks themselves show the manner in which the pre-existing Roman road was made use of by the makers of Wansdyke. The road, which is on a terrace five or six yards wide along the steep north slope of the down, is joined on the south by Wansdyke, here a deep ditch with a high rampart on the south of it, together some 27 or 30 yards wide, at the head of a coombe about a quarter of a mile east of the Calne road; and there the whole character of the intrenchment changes. There is no longer a rampart on the south of the ditch, but the edge of the embanked terrace has been raised by material excavated from a ditch along the road, and thrown up as a counterscarp some four feet high on the north of the ditch, the steep slope of the down rendering any other defence unnecessary, and the whole is not more than 17 yards wide. It is probable that the rampart crossed the road and joined the counterscarp, but that is now obliterated by a cart-track. As  p334 the slope of the hill-side gets less steep going westward, a rampart appears on the south side and increases in height, and near Smallgrain plantation, to the east of the Calne and Devizes road, there is a considerable bank on either side of a hollow way.

From the junction of Wansdyke the course of the Roman road is in a straight line between the top of Morgan's Hill (847′) and the south end of Lansdowne (760′), 18 miles distant; and it may be seen from the former to continue in that direction for three miles to near Wans. Beyond the Calne and Devizes road the ground falls, and cultivated land is entered upon, and a hedgerow followed by a parish boundary marks the line, and there is little other trace of road or dyke for one and a quarter miles, when it appears more plainly for half-a‑mile to the south of Stockley. Sir R. Colt Hoare mentions hereabouts traces of a ridge with a bank and ditch on the north side. Hedges followed by parish boundaries continue the line, and looking back from near Wans a straight line of hedges can be seen rising out of the lower ground towards Morgan's Hill. On nearing Wans the hedgerow and parish boundary bend towards the north, and soon the latter is the only trace now left of the Roman road; though Sir R. Colt Hoare's map (1819) shows the ridge to within one-eighth of a mile of the lane, which the parish boundary joins about 60 yards south of the cross roads, and follows across the Calne road. Then the ridge of the Roman road is plain in the belt of trees on the east of the grounds of Wans  p335 House, with the remains of a ditch on the north-east side, along which the parish boundary runs. The ridge bends round to the house and there it is effaced for 200 yards; but the parish boundary marks the line of it on to the Chippenham road (475′). Then another straight line is entered upon pointing to high ground (600′), south of Ashley Wood, 10 miles distant, over­looking the Avon valley near Bathford. This line is nearly parallel with that from Morgan's Hill, and a quarter of a mile to the north of it, and is joined by a rough reversed S curve. Sir R. Colt Hoare placed Verlucio at Wans, where much Roman pottery and other objects have been found, and thought that the road entered the station at the south-east angle and left it at the north-west angle. There are no remains of a camp, and there is no reason to suppose that the course of the road was affected by the station, which there is some reason to think lay to the north of Wans House. It would rather seem that, as at Elstree on Watling Street, two high points were made use of in laying out the road, the second being perhaps Calston Down (800′), near Oldbury, a mile north-east of Morgan's Hill, to which the straight line from Ashley Wood points.

At the Chippenham road Spye Park is entered, and for about 300 yards, where the ground was formerly ploughed, a low undulation of the surface is all that remains of the ridge. It then appears plain for a quarter of a mile; five or six yards across the top and four or five feet high, having on the north side traces of a ditch with a ramp or counterscarp outside, showing  p336 the modification of the road by the makers of Wansdyke. According to a section made here by Sir R. C. Hoare's surveyor in 1820, "the agger was 20 feet broad and six feet high, and the smaller one rising on the outer side of the ditch nine feet wide."​5 A steep-sided valley appears to have been crossed by winding up stream and back again to the same line, as in many other instances, and on the west side of the valley the banking up of the road on the low side is plain, and there seem to be remains of the road surface on both sides of the valley. The ridge continues for a quarter of a mile on to another valley, beyond which there are now no further traces within the park, though the ridge is shown on the old Ordnance map, and on Sir R. Colt Hoare's map (1819). There are some indications of the ridge outside the park, and in less than a mile a line of hedgerows, with remains of the ridge here and there, and followed by parish boundaries, takes up the same straight line for upwards of eight miles. From Bowden Hill (500′) the entire course of the road to Ashley Wood is in sight. Sir R. Colt Hoare noticed that a cottage built upon the Roman road at Forest Gate was a free tenement,​6 a similar case to that already mentioned on the Foss road near Radstock (p263). On the east of the canal near Lacock he describes the ridge as being 20 feet wide and four feet high with a regular trench on the north side. Where the Roman road is crossed by the railway one and a half miles north of Melksham, the  p337 parish boundary leaves the line of the road and no trace appears for a quarter of a mile; and then the course of the road, followed by a parish boundary, curves round a low hill as if to avoid the highest part, and resumes the same straight line. Near Neston Park Sir R. Colt Hoare noticed the agger six feet high with a ditch on the north side, and in the park he described a section as being four paces wide and six and a half feet high, at top a layer of loose stones, then one of earth, lower down a stratum of stone grouted or pounded,​7 the lowest foundation being concealed. It would seem from a note on the authority of Mr. Leaman that "the bank and ditch of Wansdyke were there plainly visible, as made on the foundation of the previous Roman road."​8 At Wraxhall Wood, about two miles further on, the remains of the road are described by Sir R. Colt Hoare as 30 feet wide and nearly 12 feet high,​9 and he says that the flat stones which formed the foundation of the road, with a concrete layer over them, were still to be seen. There is now to be seen a somewhat dilapidated fence wall of rough local stone, of which there is plenty about, upon what appears to be the bank of Wansdyke on the north of the track of the Roman road. Through the arable land beyond an undulation of the surface is all that remains, which would hardly be noticed if it were not in the line of the road marked by a parish boundary. There seem to be some traces of the Roman road with Wansdyke  p338 alongside it towards Ashley Wood, in which Sir R. C. Hoare found the ridge visible at the beginning of the nineteenth century.

On the high ground above Bathford, Wansdyke leaves the Roman road and turns south-west, and the course of the latter through Bathford is probably along the line of Chapel Row and round the bend of the Avon to Batheaston, where it is joined by the Foss road, and thence with a sharp turn, straight for two miles to Walcot, where the road into the Roman city of Aquae Solis (or Sulis) probably turned off. The road followed by Iter XIV went on through Weston and by North Stoke to Bitton.

The course of the road from Bath westward has been the subject of much difference of opinion. Iter XIV of Antonine's Itinerary gives the distances of the stations, beginning from Aquae Solis, thus:

To Trajectus vi M.P.
  " Abona ix M.P.
  " Venta Silurum xiv M.P.
  " Isca ix M.P.

At six miles from Bath is Bitton, where there are traces of a Roman camp and evidences of a station, but only an insignificant stream, the Boyd, to be forded, and to give the name Trajectus to a station; and there has been a disposition to assume that Trajectus is connected with a ferry across the Severn. Harris,​10 and others following him, placed Trajectus at Aust, 20 miles from Bath, deriving that name from  p339 Augustus, and Gale put it at Oldbury, three miles still higher up the Severn. Bishop Clifford​11 declined to throw over the Itinerary, and supposed that the name Trajectus has reference to a passage across the Avon near Bitton, by a British road going north and south. Coxe,​12 writing in 1801, continued the road from Bitton to St. George's Church by the south of Redland Down, and over Durdham Down, where he says it was still high and visible, to Sea Mills, near the confluence of the river Trim with the Avon, three and a half miles from the mouth of the latter; and there he places Abona. Thence he says that paved remains of the road still existed, joining the turnpike road, and so on to the Severn. On Durdham Down, the ridge still appears as a low undulation in the surface. It was opened on the west of the Stoke road, near Durdham Lodge, in 1900. Under three inches​13 of turf a paving of large stones was found, bedded on a layer of about six inches of earth, under which was about a foot of calcareous earth and limestone. Traces of the side ditches were found, about 16 yards apart, and one foot two inches to one foot nine inches below the crown of the paving of the road, which appears to have been about 20 feet wide. The distance from Sea Mills to Venta Silurum by Southbrook is rather more than thirteen miles, compared with the 14 M.P. of the Itinerary.

 p340  Bishop Clifford objected that vessels going down the Avon from Sea Mills with the tide would have the ebb against them in the Severn going up to Southbrook, and places Abona near King's Weston Park, with a landing-place on the Severn. There is no doubt that in Roman times the tide flowed up to the foot of the high ground on which he places Abona. It would do so now if it were not banked out, the land for a width of one and a half miles bordering on the Severn being several feet below high-water level. The Bishop suggests a causeway​14 and mentions Chittening Street, which is three miles from the mouth of the Avon, as probably representing it. Another way to the Severn is marked "Roman road" on the new Ordnance map. But such a causeway appears an insufficient provision for crossing a wide river with a 40 feet tide without a sheltered port at hand such as Abona at Sea Mills would have afforded. From the mouth of the Avon there is a straight run of about five miles in a slanting direction across the Severn to Portskewet, below "The Shoots," and clear of rocks, and by choosing the ebb or flow a favourable tide in either direction could be secured for the passage. Waiting for a change of tide would not have been a serious matter, and a causeway may have been available at favourable times.

On the Welsh side of the Severn a landing-place may have existed near the Roman camp at Sudbrook, of which now only half remains on the cliff near the  p341 Severn Tunnel pumping station. Shelter was certainly afforded by St. Pierre Pill and Caldicot Pill on either side of the camp, and about two and a half miles apart. The former is the outfall of the Meyric stream, and Caldicot Pill is the estuary of a stream originally the Twrch, Latinized as Tarocus, and now called the Troggy. Both pills were used by shipping until the construction of the railway, and they were so used from very early times. A grant, preserved in the Liber Llandavensis, of the land lying to the west of St. Pierre Pill, then called Pwll Meyric, includes the free right to bring ships to land in the mouth of the Troggy (in hostio Taroci), i.e. in Caldicot Pill. The grants probably date from the seventh century, and the conclusion is that at that time both were old-established harbours.15

For the passage to and from the Avon, Caldicot Pill would be the more convenient harbour, and a paved causeway has been traced from Caldicot to Caerwent. It enters Caldicot common field, where it is soon lost under the soil; it is marked for a short length as "track of Portway" on the Ordnance map, and is joined by the present hauling way for about half-a‑mile, and is then lost as far as the Nedern brook.​16 Coxe in 1801 found vestiges of an ancient paved causeway between the brook and the eastern  p342 gate of Caerwent, which, within the memory of man, had been more perfect.17

The Roman station Venta Silurum at Caerwent measures about 530 yards from east to west, and 450 yards from north to south. The road runs through the middle lengthways, and on straight for upwards of a mile, followed for three-quarters of a mile by a parish boundary, to Five Lanes, where there were vestiges of a causeway in 1801.​18 The present road then turns to the north, but in a mile and a half resumes the same line near Penhow. Two miles further on, at Catash, the course of the Roman road appears to follow a parish boundary on the north of the present road. It leaves the latter again near Woodville, and passing by Summerhill, crosses the Usk and enters Caerleon in the line of High Street.

The rectangle enclosed by the walls of the Roman Isca leg. II Augusta, which can still be traced, measures about 530 yards by 450 yards. It is thus of exactly the same dimensions as Venta. This station, the Isca of the Itinerary, is the starting-point of Iter XIV, which has now been followed in the reverse direction from Silchester, and also of Iter XIII by Burrium to Gloucester, Speen, and Silchester. Burrium is placed near Usk, the distances from Caerleon, and from Abergavenny, Gobannium in Iter XII, corresponding. There appears to be no trace of the road from Caerleon to Usk, nor of a Roman road onwards in the direction of Gloucester.  p343 The indications of a Roman road on to Abergavenny are only a few short lengths of boundary along the present road, and no Roman remains are known at Abergavenny.

Camden 19 cites a passage quoted by Leland from a poem by Necham (1215‑1225), in which Julia Strata is mentioned in connexion with the Usk entering the Severn.​20 On this hint the author of the spurious Itinerary of Richard of Cirencester gave the name Via Julia to his Iter from Bath to St. Davids. Sir R. Colt Hoare made it Via Julia Maritima, and Scarth extended it eastwards to Marlborough, Speen, and Silchester. The only authority for this name is the reference above mentioned.

On the west of Caerleon, on the north side of the railway, traces of the Roman road are marked on the Ordnance map, which also marks the road from Bassaleg onwards as a Roman road. About one mile from Bassaleg traces of the ridge remain, and the road appears to have been laid out straight for one and a half miles. A little further on is a square camp by the side of the road, one mile beyond which the modern road turns south  p344 to St. Mellon's, Rumney, and Cardiff.

Cardiff Castle has generally been assumed to have been where the Roman road crossed the river Tarr; but it is probable that the old road continued straight on where the modern road turns south to St. Mellon's, a line of roads from Julian's Farm a mile and a half further on, and parish boundaries for a mile on the north of the Taff at Gabalva opposite Landaff, and for a quarter of a mile in the same line north of Ely Bridge, indicating the line of the Roman road.

There is no doubt, however, that a Roman fortress guarded the mouth of the river Taff two miles lower down, at Cardiff Castle. The Roman walls have of late years been discovered beneath the mediaeval ramparts along the eastern and northern and part of the western sides; the Castle buildings covering the rest of the latter, and the southern side. The rectangle of the Roman walls measures 198 yards from east to west, and about 210 yards from north to south, with a gateway in the northern face.21

Beyond Ely Bridge the modern road to Cowbridge is followed by parish boundaries for half-a‑mile near Caerau, where there are remains of a Roman camp about half-a‑mile to the south of the road, and after a break, again for a mile and a quarter, where the boundary follows the old road up the hill to St. Lythan's Down, which the modern road goes round. On the west of the down the parish boundary again follows the road over the hill, and continues along the modern road for a quarter of a mile along Coedriglan Park. There are other indications which seem to show that the present road follows the line of the Roman road through Bonvilston, deviating from it at Llantrythid Park, between Old Post and Three Ashes, and again at  p345 Stalling Down. Here the characteristic features of a Roman road reappear. From Stalling Down (400′) one mile east of Cowbridge, to Stormy Down (310′) 11¼ miles distant, there lie in a straight line half-a‑mile of highway, half-a‑mile of Cowbridge street, and then three and a half miles of the main road as far as Brocastle. The line is continued on by a lane in the direction of Ewenny Abbey for a mile, and then there is no trace for two and a half miles across the Ewenny and Ogmore valley. A lane within a parish boundary along it for a quarter of a mile then takes up the same line for three-quarters of a mile on the south of Laleston. After an interval of a mile the same line is joined by the main road, which follows it for a mile, with a parish boundary along it from Park Isaf to Twmpath-y‑ddar, where there is a Roman camp (331′) and a clear view along the same line back to Stalling Down. The main road and the parish boundary turn off northward to Pyle, but the line of the Roman road continues on with visible traces in nearly the same line by Heol-y‑sheel, and Heol-las, to Kenfig. From Kenfig the road seems to have turned north along Water Street, past what is marked as a Roman monument on the old Ordnance map, to Beggarsbush on the main road near Margam. It possibly continued on over the Margam mountain, and by the east of Mynydd-y‑gaer crowned by a camp, to join Sarn Helen. This would be more in accordance with Roman practice than a course along the shore of the Neath estuary, commanded on the land side by high and steep hills.  p346 The difficulties of the way across the sands from Margam to Neath at the end of the twelfth century are described by Giraldus Cambrensis. The finding of two milliaries, near Port Talbot and near Aberavon, may be thought to show that the road passed those places, but the original positions of the stones are unknown. No traces of the road are known, nor are there any vestiges of a Roman station, nor of Roman remains of any sort at or near Neath.

Iter XII of Antonine has been thought to pass over the road which has now been followed. Camden, who supposed that transcribers had "strangely confounded the two Iters from Calleva to Isca, and from Maridunum to Viriconium," placed Bomium near Cowbridge, Nidum at Neath, Leucarum at Loughor (Llwchwr), and Muridunum (Maridunum) at Carmarthen. In this he was followed by the writers of the spurious Itinerary of Richard of Cirencester, and by Bishop Bennet and others. If Parthey and Pinder are right in supposing that the first eight lines of Iter XII were transferred by mistake from Iter XV, the next station to Muridunum would be Leucarum, XV M.P. distant, compared with 19 miles from Carmarthen to Loughor; the XV M.P. on to Nidum would compare with 13 miles on to Neath; and the XV M.P. on to Bomium, with 22 miles to a point near Cowbridge, 27 miles (the Itinerary distance) from Caerleon. Neither the total nor the intermediate distances agree; no traces of Roman stations are known at Loughor, Neath, or near Cowbridge, although an altar, coins and pottery have been found at  p347 Loughor; and there are no evidences of a Roman road between Carmarthen and Neath. Muridunum in Devonshire appears in Iter XV, and Ridunum in the Tabula Peutingeriana, in both XV M.P. from Isca Dumnuniorum; while Maridunum of Ptolemy is stated by him to be in the country of the Dimetae, that is to say in South Wales. Whether the latter is the Muridunum mentioned in Iter XII must remain uncertain.

(11) Speen to Gloucester

There is no trace of this road until the gravel-covered tertiary beds are quitted near Wickham, four and a half miles from Speen. There a road is entered upon which is straight in general direction for nine miles to Baydon Hill (786′), and points in the other direction to Speen. For some miles from Wickham the present road is not straight, being sometimes on the line of the Roman road, but generally winding from one side of it to the other. There is little trace of the ridge until beyond Poughlye, where a piece of it remains on the north of the present road, which there takes the line of the Roman road, and by its elevation between hedges wide apart shows traces of the ridge. The road is soon narrowed to 12 or 15 feet between the fences, and the Hare and Hounds Inn stands on the line of the Roman road. At Woodlands Farm the present road turns south at a right angle, and the course of the Roman road can be traced straight on across a meadow by an undulation of the surface, with browner herbage. In the arable field beyond the stones of the road were ploughed up  p348 20 to 25 years ago. In three-quarters of a mile the present road rejoins the Roman road and follows it through Baydon. In the beginning of the nineteenth century the Roman road hereabouts is described by Bishop Bennet​22 as presenting an elevated crest raised many feet above the downs in various parts. Beyond Baydon a parish boundary follows the present road for a mile, and continues on in the same straight line for three-quarters of a mile further to Peaks Down, and then the Roman road is traceable across the down in the direction of Wanborough Plain Farm. Near this place remains have been found, which may mark the site of the Roman station already mentioned (p328) as having dropped out of Iter XIII between Durocornovium and Spinae. The road thence turns due north-west, and runs straight for Callas Hill on the escarpment of the chalk half-a‑mile to the east of Wanborough, and then a straight road is entered upon which passes through Stratton St. Margaret's to Blunsdon Hill (494′), seven miles distant. On this length of road, a mile and a half from Stratton St. Margaret's, and three miles due east of Swindon station, is the junction of the Roman road from Winchester, and here Sir R. C. Hoare placed a Roman station to which he gave the name of Nidum, which has already been referred to (p328). Looking back from Blunsdon Hill the road can be seen to the south-east mounting up Callas Hill, the chalk downs beyond being visible; and towards the north-west Cricklade and Cirencester, the latter ten miles off, with  p349 the Cotswold hills beyond, are in sight. The road, where it has not been encroached upon, is 50 or 60 feet between the fences, but it has generally been narrowed on one side or the other, and in consequence the straightness is not so apparent. On Blunsdon Hill there is a slight turn, and a straight road runs to Calcott Bridge, near Cricklade. There is now an interval of nearly a mile at Cricklade, in which, however, portions of a causeway across the meadows were dug up at the end of the eighteenth century.​23 Parish boundaries run along the road from Blunsdon to Seven Bridges, Water Easton.

From the north-west of Cricklade a straight road runs for three and a half miles to Driffield Cross (320′); looking back from which Blunsdon Hill is plain, and the road up the chalk escarpment at Callas Hill can be seen in clear weather. Near Driffield Cross there is a round turn and then another straight road leads to Cirencester. This straight line points from Driffield Cross, to the south-east gate of Cirencester, and to high ground (700′) near Duntisbourne House, four miles north-west of Cirencester. If it were not for the elm trees along the road Driffield Cross would be visible from the entrance of Cirencester. Parish boundaries follow the road for 2 miles and three-quarters through Driffield Cross to Siddington. The road is not much raised above the surface until Preston is reached, and then for about half-a‑mile it is carried across the meadows to Cirencester, on a causeway on the south side of the river Churn.

 p350  From Wanborough Plain Farm, where there is a decided change in the general direction, to Cirencester, no part of the road is as much as half-a‑mile away from an absolutely straight line 19 miles long. The road here bears the name of the Ermin Way, as it is called by Higden, the Monk of Chester, and under that name it goes on through Cirencester to Gloucester. It constituted the main street of the Roman Corinium, or Durocornovium, which, between the south-east gate and the church, is now represented for 150 yards only by Tower Street, the rest being built over. North of the church, Dollar Street and Gloucester Street occupy the site of the same Roman street, which measures seven furlongs from gate to gate. About midway it was crossed at right angles by the principal street about half-a‑mile long from gate to gate, now represented by Lewis Lane and Queen's Lane, and the other Roman streets and buildings so far as they are known appear to be laid out parallel to, and at right angles with Ermin Way. The enclosing walls, which are still traceable, do not suggest the rectangular plan of a Roman camp; and they enclose an area far larger than the largest Roman camps such as Glevum, Venta Silurum, or Isca. In these respects Cirencester resembles Verulam, Colchester, Chichester, and Silchester; British cities rebuilt by the Romans, and afterwards walled.

From Cirencester the course of the Roman road, following the present road, is straight for eight miles to Gloucester Beeches (900′), with a slight turn on high ground (500′) at Daglingworth Down, and with  p351 some twisting up the hill by Stratton, due in part at any rate to modern improvement of the gradient. The ridge of the Roman road makes its appearance soon after leaving Cirencester, and on the high ground it is now seven yards wide and four to five feet high, and near the third milestone as much as six feet high. There are ditches at the sides in places which may be in part Roman, or may be more recent quarries for road materials. Looking back from this high ground the chalk downs, which are crossed by the Roman road, 20 to 22 miles off, to the south-east, are visible. Beyond Gloucester Beeches the road passes round the head of the Side valley at Nettlecombe to Birdlip, and winds down the hill, followed by parish boundaries. Near Little Witcombe, five miles from Gloucester, a straight line is entered upon which, as one looks back from Wootton Cross roads one mile east of the cathedral, points to high ground (900′) above Birdlip Hill. Stukeley says that from Cirencester the Roman road appeared with a very high ridge and very straight for eight miles to Birdlip Hill​24 and Camden says that on Birdlip Hill it showed a very bold ridge and appeared to have been paved with stone.​25 Where the road is crossed by the railway the Roman paving was found entire 18 inches beneath the surface. The road turned to the south at Wootton Cross and entered Clevum (or Glevum) by the north gate, together with the road from the north. In Northgate Street, Gloucester, the  p352 Roman road was found about ten feet six inches below the surface, composed of stones of irregular shape bedded in mortar on concrete.

The walls of the Roman Clevum, as traced by Bellows,​26 included a rectangle 510 yards by 435 yards. The north corner was in the Cathedral cloisters, the east corner at the junction of Aldgate Street and King Street, the south corner inside the turn from Brunswick Road into Parliament Street, and the west corner at the angle of the County Prison facing Commercial Road.

The road from Gloucester to the north appears to have been on the line of the Tewkesbury road passing through Longford, where a closely-pitched pavement was found six feet below the surface of the road. To the north of Twigworth, a parish boundary follows the present road for about a mile, and beyond that there seem to be no indications of the course for more than nine miles, and then Stratford, at the boundary of Gloucestershire, seems to mark the line. From about four miles north of Stratford, highways, tracks, footways, and a parish boundary continue in a line between Stratford and Worcester for more than six miles. A footway crosses Croome Park from Kinnersley to a highway in the same line for half-a‑mile, and then a footway leads to and follows a belt of old trees in the same line for half-a‑mile, followed for a quarter of a mile by a parish boundary. A narrow strip of enclosed land on the west of Kempsey Common, with the footway alongside, continues the  p353 line to a broad lane leading on to Napleton, where the straight course is lost for a quarter of a mile. A lane continues it for a mile, and then a footway takes up the line for a mile and a quarter, passing on the east of Timberdine Farm, to the Worcester road. After crossing the latter the course is across pasture fields, marked by a line of old trees, to a footbridge over Duck Pool, and then a highway takes up the line, pointing in the direction of High Street, Worcester.

There is no definite trace of a ridge along this line, but there are appearances which suggest that the Roman road materials have been removed, leaving a shallow hollow along the course of the road. The same thing is to be seen along the course of Riknild Street through Hindlip Park, on the north of Worcester (p271); and several other instances where the ridge has been removed, and the materials have been dug out, leaving a shallow trench, have been already noticed.

Parish boundaries along the present road from Gloucester towards Bristol, at Hempsted, Quedgeley, Moreton Valence, and on the south of Stroud Water, for lengths of a quarter of a mile to a mile and a half, seem to indicate the line of the Roman road in that direction.

From the west gate of Clevum the Roman road proceeded in a north-westerly direction straight to the Severn. The causeway, raised some five feet above the meadows, now called Over Causeway, is undoubtedly in part Roman. An old man lately described it as having been paved in the middle with  p354 cobble stones in his father's time, the width of the road being then the same as it is now, about 10 yards. The approach to Telford's bridge at Over turns off to the north, and the causeway beyond has been destroyed, but a bit of the old road with the narrow bridge over the old course of the river Leadon, in line with the causeway, is still to be seen on the west of the river between the approach to Over bridge and the railway. The Ordnance maps mark a road along the low eastern bank of the Severn northward to Maisemore Bridge as a Roman road, on what evidence is not apparent.

(12) Gloucester to Redbrook, and to Caerwent

On the west of the Severn there is some indication that the Roman road onwards took a direct course from Over to Linton, from which the present road points for three and a half miles to high ground (630′) to the south of Mitcheldean. Roman roads are supposed to have branched from near Over both to the south-west and the north-west. The course of the former, of which there seem to be no traces, is marked on the Ordnance map through the meadows and by Murcot to Minsterworth, falling into the present road from Gloucester to Westbury and Newnham. The latter road is thereabouts below the level of the Severn floods, and a little further on is connected by Oakle Street with the road westward from Gloucester. It is marked on the Ordnance map as a Roman road, and "traces of Roman paving" are marked in two places on it further on, but it has none of the features of Roman laying out. It is an old road, which was  p355 a turnpike road for many years, and was improved by Telford.

"Traces of Roman paving" are marked in many places of the Ordnance maps on the roads of this part of the country. They are to be seen on roads which were improved as turnpike roads in the latter part of the eighteenth and the early part of the nineteenth century, and which were in some cases afterwards improved as mail roads. The traces are marked where the roads have been diverted or lowered to improve the gradients, and also where the roads are sunk in the ground as if from the effects of long wear previous to the laying of the paving. Paving was no doubt to be seen where it is marked, and traces of it, or more frequently of a line of kerbing, are still visible here and there, but it by no means follows that it was always Roman. In the Forest of Dean, paved Roman roads still remain, and on old roads and tracks which have lately been modernized, the traces may in some cases have been of Roman paving.​27 But when roads were improved in the early part of the last century, and if repairs  p356 were afterwards neglected, the road wore away, and the paving stones became exposed on the surface. This was the case in 1881 on the Shrewsbury and Holyhead road, constructed with a paved foundation in 1823, and may well have been the case in the neighbourhood now under notice, where the roads were much worse cared for.

From Newnham onwards to beyond Blakeney many "traces of Roman paving," now no longer visible, are marked along the road on the Ordnance maps, but there is nothing characteristic of a Roman road. About a mile south of Blakeney, the road takes a more direct course, and follows the same general direction to Lydney, where it is joined by the Dean road, the connexion of which with Gloucester must be now traced.

The present road from Over, which points for three miles and a half from Linton to high ground on the south of Mitcheldean, is followed by parish boundaries for a mile and a quarter, and with some improvements of gradient appears to follow the course of the Roman road to beyond Halfway Bridge. It there turns towards the north-west, but the direct line is continued on by a lane, a footway and parish boundaries for two and a half miles, and then the present road by Huntley rejoins, and follows the direct line onwards for a mile. This is the course followed by the post road from Gloucester to Mitcheldean in Overton's map of 1712, which appears to be the earliest map on which roads are shown. The course of the Roman road on towards Mitcheldean is uncertain.  p357 The post road on the old maps continues on from Mitcheldean by Drybrook to Monmouth, and "traces of Roman paving" are marked on the Ordnance map at many points on the road onwards between Mitcheldean and Drybrook, and it is marked as a Roman road on to Ruardean; and beyond "traces of Roman paving" are marked along a narrow road at Lower Lydbrook, near the Wye.

At or near Mitcheldean a Roman road turned to the south, but the course is uncertain for a mile and a half. At three-quarters of a mile south of Mitcheldean, the Ordnance map marks "traces of Roman paving" for half-a‑mile beyond the turn in the present road near Abinghall, but it seems likely that a lane along the higher ground from the Folly to near Gun's Mill may be the course of the Roman road. To the south of Gun's Mill, the paving, consisting of squared blocks, and apparently Roman, was entire not many years ago. The Ordnance maps mark "traces of Roman paving" in six places in the mile and a half onwards by Tibbs Cross to Littledean, and the kerb bordering is still to be seen here and there in the improved road.

At Littledean two roads from the east converge. They are for the most part narrow, sunk, crooked roads, with little to suggest a Roman origin, but both are marked all along on the Ordnance map with "traces of Roman paving." On the road which passes from the Severn side through Newnham to the south of Littledean, a paving of squared blocks is said to have been entire not many years ago, but  p358 some of the "traces of Roman paving" are marked on the maps in places where the road must have been worn into a deep hollow before the paving was laid. Westward from Littledean, in continuation of this road, "traces of Roman paving" are marked on the Ordnance maps in eight places in the mile and a quarter to the Speech House, but with the recent improvement of the road, all have now gone. On the hill between the Speech House and the railway-station, a strip of bordering kerbstones and some of the paving, said to be Roman, remained not long ago. A paved Roman road in the same direction still remains in the Forest to the west of the railway, and to the north of the present road, and it can be traced for some distance pointing towards Worcester Lodge. It is of squared stones bordered with kerbing, and about eight feet wide. It may have joined a Roman road along the course of the old post road from Mitcheldean, which seems to be that of a road mentioned in a perambulation of the bailiwicks of the Forest in the tenth year of Edward I, as altam viam tendens apud Monmouth.​28 The course onwards is by Coleford Meend and Lower Berry Hill to Cherry Orchard near Newland railway-station, where a road from Clearwell through Scatterford and Newland joins. Traces of Roman paving are marked along the latter on the Ordnance maps, and Mr. J. G. Wood describes it and the kerbing as being still visible after rain, and plainly Roman. This road is in continuation of a  p359 straight road to Bream Cross in the direction of Lydney Park, two and a half miles distant.

A parish boundary, and then a county boundary, follows the road from Cherry Orchard down to the Wye at Redbrook, where probably the river was crossed to reach Blestium. The old post road turns off to the Wye bridge at Monmouth half-a‑mile from Redbrook, passing by Duffield Lane to the south-west of the Kymin.

The Ordnance map marks as a Roman road a track in the wood on the north side of the road from Staunton to Monmouth, which a county boundary follows for half-a‑mile from Staunton. The track is crossed twice by the modern road, and can then be followed as a narrow, hollow way northwards to the river Wye at Hadnock, a mile and three-quarters above Monmouth. It has the appearance of an old packhorse road, which however may have been in use in town rimes, and even earlier. Roman coins have been found in the workings of the iron mines near Clearwell, and coins, fibulae, and other Roman objects have been met with deep in the cinder-heaps of the old bloomeries, especially at Whitchurch, on the west of the Wye, two miles to the north of Hadnock; showing that the conveyance of the haematite ore for reduction to bloomeries, situated where wood fuel was plentiful, went on at least as early as the Roman occupation.

Continuing southward from Littledean, traces of paving are still to be seen in the sunk road near Grange; and along the present modernized road  p360 in Sutton Bottom "traces of Roman paving" are marked in six places on the Ordnance map. At about half-a‑mile from Soudley the present road diverges from the old track, and along the latter traces of paving can still be seen. It soon divides — one track going straight to Soudley Camp, and the other through Soudley Plantation to the Dean road at the ford. Soudley Camp occupies the end of a ridge between two brooks; it is triangular in shape, each side measuring about 60 yards. From Soudley Ford the paving of the Roman road is plain, and onwards traces and remains of a paved road distinctly Roman are almost continuous along the forest track called Dean Road. Towards Blackpool Bridge the pavement remains almost perfect for 150 yards. It consists of blocks, about eight inches square, between kerbstones five or six inches wide, and 10 to 18 inches long. In some places the kerbstones are held in place by blocks of stone outside, breaking joint with them, and this seems to be the case where a less massive kerbstone has been used. Towards Soudley the width is as much as nine feet, but the more perfect length towards Blackpool is eight feet wide. It can be traced on both sides of the stream at Blackpool to the west of the bridge. The bridge is not 100 years old, but it may be noticed that there are plain traces of paving leading to it, as well as to the ford. The Roman paving is again plainly visible on the Dean road for 40 or 50 yards about 100 yards south of the Blakeney road, where it is nine feet wide. Neither this nor the Roman paved road to  p361 the west of the railway, near Speech House station, is noticed on the Ordnance maps. The Dean road continues on as a green track through the forest for another mile and a half to Old Croft, and on to Allaston, and then, following a track and footway, it falls into the main road from Gloucester on the north of Newerne.

In the perambulation of the bailiwicks already referred to, Dean Road is called via Regia.

It has already been said (p356) that from about a mile south of Bakeney the present road from Gloucester takes a more direct course. The same general direction is followed from The Purlieu past the point of junction of the Dean road, and through Newerne and Lydney, and then a footway by Lydney House follows the same line to Aylburton, where the present road takes it up.

In Lydney Park many Roman remains, the foundation of a temple, and of a large villa, with baths, pavements, statues and objects in bronze and iron, show it to have been a place of some importance.

The Ordnance maps mark "traces of Roman paving" in the present road on the south of Aylburton, but the road seems to have been altered by modern improvements, and it may be doubted if the paving, now no longer visible, was Roman. It appears probable that the old road continued straight on by Sandford Bridge to Alvington. On the south of Alvington, after crossing the brook, the present road turns to the eastward, but a track and a footway  p362 straight on up the hill continues the line to join the present road near Brookend, and seems to be the old course. The road follows the same line onwards for a mile, except at a twist where the stream at Wyeford is crossed, and then there is a slight turn on high ground, and a parish boundary follows the road for three-quarters of a mile to Stroat, where a slight deviation seems to have been made. The present road continues straight in general direction to Tidenham, and then bends towards the west, and about a mile beyond Tidenham the course of the Roman road leaves the present road, and is followed by a footway to the north of Tutshill. In 1861 the road showed its line through the turf, and was found as a rude pavement in sinking foundations for buildings.​29 It could be traced down to the Wye at the site of the ancient bridge across the river at the south end of Piercefield Park, where remains of piles were clearly to be seen at low water. The track can still be traced on the Monmouthshire shore, ascending through Castle Wood to The Mount, Chepstow, from which a parish boundary and a road mark the line to Hardwick Hill. The parish boundary turns along the Newport road for more than a quarter of a mile, and then the Ordnance map marks "traces of Roman paving" on it. In about half-a‑mile the road passes through Pwll Meyric, and on to New Inn Smithy, where it bends to the south, and the course of the Roman road continues on to a highway leading to Crick and Caerwent. In 1732 Horsley described it on  p363 the east of Caerwent as a military way large and remarkable.30

(13) Iter XIII of Antonine

The present road from Gloucester which turns from the line of the old post road towards the north-west at about five miles from Gloucester, is marked as a Roman road on the Ordnance maps with "traces of Roman paving" along it. The road was improved by Telford as the mail road to Ireland by way of Milford Haven in the earlier part of the last century, and it has no appearance of Roman laying out. It is crossed by the road from Mitcheldean to Newent, marked as a Roman road on the Ordnance map, and with "traces of Roman paving" for some little distance. Another road northwards from Mitcheldean, by Mitcheldean Road railway-station to Brooms Ash, seems to have quite as much claim to be considered a Roman road. From it the Ordnance map marks a Roman road branching to Bury Hill near Bollitree, where Roman remains have been found. They consist mainly of coins and fibulae; and it has been thought that there are traces of hardware manufacture.​31 It has been supposed to be the site of Ariconium, a station in Iter XIII of Antonine's Itinerary, XV M.P. from Clevum; Blestium, the next station westward, XI M.P. distant, being placed at or near Monmouth. Bollitree is 13 miles from Gloucester, and 13 miles from Monmouth by way of Goodrich Ferry, and there is little evidence of a Roman road either from  p364 Gloucester or on to Monmouth, where no Roman remains are known. The same may be said of Usk, where Burrium, the next station in the Iter, has been placed; but Burrium is found also in Iter XII, and the distances from Gobannium at Abergavenny in that Iter, and from Isca at Caerleon in Iter XIII, accord in fixing Burrium at or near Usk.

A route for Iter XIII by Mitcheldean and Redbrook suggests itself as an alternative. Ariconium would then be near Littledean, and Blestium near the Wye.

(14) Gloucester to Llandovery

The Roman road appears to have turned off to the north-west after the Severn was crossed, but there is little trace of it. In a perambulation of the Forest of Dean, A.D. 1228, the road from Leadon bridge at Over to beyond Newent is called "Magna Strata"; and in a perambulation A.D. 1300, it is called "regalis via";​32 names which furnish a presumption that it was a Roman road, though it is doubtful how far the present road follows the same course. Beyond Dymoke there is more of the character of Roman setting out, the course of the present road lying in straight lines from point to point in very nearly the same direction for nine miles and a half to Stretton Grandison. The road must have fallen into the continuation of Riknild Street coming from Worcester, of which there are unfortunately no traces. From Stretton Grandison the course is perhaps straight to the present Hereford road near Shucknall, then following the old road to Lugg Bridge, and  p365 thence along a highway followed by the boundary of the city of Hereford for three and a half miles; and on very nearly in the same straight line to the station Magnae at Kenchester, parish boundaries running along the present road for most of the way. At Magnae, the road coming southward from Wroxeter, which has already been noticed (p80), crossed. Westward from Magnae the course appears to be followed by the present road as far as Garnons, and a parish boundary across the Lawns in the same general direction probably represents the line on to the Hay road to the west of Garnons. From Maddle Brook a parish boundary follows the straight road for nearly two miles to Staunton-on‑the‑Wye. Some five miles further on a parish boundary again follows the road for half-a‑mile between Winforton and Whitney, where the modern road crosses the Wye; but the Roman road seems to have continued on the north side of the river, and is probably represented by a lane turning south at Bronydd to the square camp near the bank of the Wye, which must have been crossed near Hay. Between Hay and Brecon there are but few indications of the course of the road. At about a mile from Hay a piece of straight road begins, which for two miles points straight to a conical hill near Three Cocks, and on it are two places called Ffordd fawr (great road). The course is perhaps followed by the modern road to beyond Bronllys. From near Pont-y‑bat‑fach, one and three-quarters miles west of Bronllys, a line of highways crossing the main road at Pen-isaf‑waun,  p366 and continuing on in the same line to within one mile of Brecon, on which the name of Ffordd fawr again occurs, may probably represent the Roman road.

This road was joined near Brecon by a Roman road from Abergavenny, of which however there is little trace. To the west of Crickhowell the course seems to be by Tretower to Pen-y‑gaer, where there are the remains of a Roman camp, and along a lane to Ty-maur, about half-a‑mile west of which at Bwlch a parish boundary joins the present main road and follows it for half-a‑mile, and continues along a lane for a mile, to the southward of Allt-yr‑Yscrin. A stone pillar, probably a milliary, dedicated to Victorinus (A.D. 265‑7) formerly stood by the side of the road at Scethrog,​33 and seems to show that the present road follows the course of the Roman road on the west of the Brecon and Merthyr railway.

West of Brecon the road called Hen-heol (Old Street) is no doubt the Roman road leading to the camp at Aberyscir on the north bank of the river Usk, on the east of the confluence of the Yscir. The camp called the Gaer, some of the masonry foundations of the walls of which remain, measures 207 yards by 140 yards. From it Roman roads went in four directions. Of Hen-heol to the east some traces remain, but the course of the others is not so plain.

The road northwards appears to have passed over  p367 Cefn Sarnau, where it seems probable that it was joined by a road branching on the north of Brecon, and passing over the hill to the east of Pen-y‑crug and by Sarnau. It crossed the Honddu at Castell Madoc, from which a parish boundary follows the present road for a mile and a half, and it then mounted the high ground on which from Post-y‑pabell for two miles a parish boundary runs along the track. The course is then uncertain. It is supposed to cross the Wye near Castell Llechryd, a camp 200 yards square close to Builth Road railway-station, and to have gone over Llandrindod Common, where traces are said to have been visible at the beginning of the last century, and over Cefnllys crossing the river Ithon, just south of Castell Collen, a large camp 133 yards square, with remains of walls of rough-hewn stone. In 1786 there were remains of a raised causeway made of large pebbles and gravel, and overgrown with grass.​34 The road is supposed to have continued on to Caersws near Newtown, Montgomery, by Caerfagu, over Camllo Hill and up the Clywedog valley, and over Bwlch-y‑Carnau, but the course is uncertain.

Sarn Helen to Neath

The Roman road to the south from Aberyscir, called Sarn Helen, seems to have crossed the Usk about half-a‑mile west of the camp, where a parish boundary runs at right angles from the river to the end of a ridge, and then along the top of the ridge and along a lane to Heol Fwt-y‑drain. The parish boundary continues along a road  p368 to the south of Mynydd Illtyd, though the Roman road is thought to be visible about a quarter of a mile further to the north. After passing over the Brecknockshire mountains, it appears to have entered the upper end of the Llia valley, and after following it for two miles to have turned south-west over the mountains, crossed the valley of the Neath river near its source, ascended the steep escarpment of the mountain, and followed Nant Hir down to Ton-y‑Castell, one mile south-east of Capel Coelbren. In 1760 the causeway here was described to be uninterrupted for a mile, a raised road 40 feet wide with a ditch on each side,​35 and it is still plainly visible. The 40 feet no doubt includes the ditches.

At Ton-y‑Castell there is a Roman camp about 166 yards square close to Sarn Helen, which here enters Glamorganshire; and then parish boundaries follow the road over the mountains between the vale of Neath and the Dulais valley for eight miles. The road remains plain on the mountains, and, as indicated by the parish boundary, it descended into the vale of Neath near Ynis-y‑Geryn, about two and a half miles from Neath, where it may have crossed the river, or may have followed the course of the present main road and crossed further down, where it is said to have been visible in the marsh at the beginning of the last century.36

From the station at Aberyscir the Roman road seems to have continued westward along the north  p369 of the river Usk, generally in the course now followed by highways, to one mile beyond Rhyd-y‑Brew, where the modern main road joins it. Hereabouts the causeway was visible in 1774.​37 At Trecastle the narrow Cwm Dwr, through which the modern road passes, was avoided, and the Roman road ascended the Trecastle mountain to the south of it. A rough track now represents what, until about 1785, was the coach road between Brecon and Llandovery. Near it in 1769 a stone supposed to be a milliary, inscribed to Posthumus, was found. On the edge of the west escarpment of the mountain (1267′) the road passes close to the remains of a large camp called Pigwn, now a good deal obliterated. The Ordnance map shows a rectangle 466 yards by 400 yards, with another 366 yards by 300 yards placed diagonally within it, and not far off a small camp 46 yards square. The road then winds down by Black-cock (where oxen used to be kept to help the coaches up the steep hill) to Hafod. The course of the road is then straight for a mile to the head of a deep valley, which it passes round, and again runs straight for a mile, and then winds down the hill by Fron. The road appears to have crossed the river Brân near Llandovery Castle.

At Llandovery, where Luentium of Ptolemy has been placed, but with little reason, four Roman roads met. From Llanfair-ar‑y‑bryn, half-a‑mile to the north of the town, where Roman remains have been found, there was formerly "a very notable Roman  p370 way of gravel and small pebbles"​38 in a north-easterly direction. The present road, straight for two miles, follows it, and about three miles further on Sir R. C. Hoare observed a causeway upon Cefn Llwydlo, which is now visible. The road is said to have crossed the river Yrfon at Glancamddwr, and to have joined the Roman road which has been described (p367) from Aberyscir to Castell Llechryd and Castell Collen.

(15) Llandovery to Llanio and Sarn Helen

The road onwards from Llandovery seems to have crossed the river Towy in the direction of the railway-station. On the north of the river the course appears to be shown by a highway leading over Bwlch Trebannan to Cynfil Cayo, and the Roman gold-mines at Gogofau, and on to a Roman road bearing the name of Sarn Helen running north. This Sarn Helen is parallel to, and 24 miles west of the continuation northward of the Sarn Helen from Neath to Aberyscir. It is followed by highways in a straight line for two and a half miles south of Bwlch Blaen-y‑corn, and crosses over the mountains, where it is very plain, to Llanfair Clydogau, and then it went on by a course which is obscure to Llanio, where at Caer Castell, Roman pottery, coins, and inscribed coins are found, and which has been supposed to be the Luentium of Ptolemy, but with perhaps less claim than Llandovery. A Roman road bearing the name of Sarn Helen can be plainly traced in a direct line due north from Llanio for nine miles, and it is visible further on at  p371 Cwmllechwydd, and again at Llwynrhyngell, a mile and a half to the north of the river Ystwith. It is supposed to have crossed the river Dovey at Penallt near Machynlleth, where Llwyd noticed a Roman fort, and a broad hard way of pitched stones from it to the waterside in a straight line for 200 yards, and 10 or 12 feet in breadth.​39 This road is supposed to have joined the Sarn Helen coming southwards from Conovium (p89).

(16) Llandovery to Carmarthen

Soon after the river Towy was crossed, a Roman road branched off at right angles to Carmarthen. It has been traced in the grounds of Blaen-nos, and about a mile further on the present main road is supposed to take the line of it, though as far as Llanwrda it lies under a steep hill in a position not usual for a Roman road. At Abermarlas the modern road turns south, but the Roman road ascended a ridge of high ground over­looking the Towy valley, and is represented by a line of highways passing by Cefn-glas‑fryn, and a mile and a half north of Llandilo. The modern road is crossed two miles west of Llandilo, and again a mile and a half further on at Broad Oak; the line of highways following the course of the Roman road rejoins the present road at Pontdulas, but the Roman road may have crossed the Dulas rather higher up the stream. There is little in the present road, narrow and crooked, though direct in its general course, to suggest a Roman road; but paving here and there may be vestiges of one, and it is locally known as the  p372 Roman road. It seems to have passed over the hill by Pen-cae‑gwyn to Halfway House, and in less than a mile an old highway on the south of the modern road takes up the line by Llanegwad to Pen-yr‑heol, from which the main road follows the course of the Roman road through Abergwili to Tanerdy near Carmarthen. It seems probable that it continued on in the same direction, skirting the north of the town, and taking the line of Catherine Street to Picton Terrace. Carmarthen is generally supposed to be Maridunum of Ptolemy, one of two towns in the country of the Dimetae of which he gives the position, the other being Luentium. According to the degrees of latitude and longitude, by which the positions are given, Luentium is about 25 miles south-east of Maridunum, so that if the latter be Carmarthen, the former cannot be at Llanio, nor at Llandovery.​c

Carmarthen to Llanio

At Tanerdy on the east of Carmarthen a Roman road appears to have branched northwards along the line now followed by the main road to Llanbyther, except where here and there the old course has been quitted to improve the gradient. It is visible in places and known as Sarn Helen. It passed through Pencarreg, and at the beginning of the last century it was visible in the bank of the river Teifi, and in the meadows adjoining near Lampeter bridge.​40 It no doubt joined the Roman road from Cynfil Cayo near Llanfair Clydogau, and with it continued on to Llanio.

(17) Carmarthen to the West

Westwards from  p373 Carmarthen the course of the Roman road appears to be that of the present main road for about three and a half miles. It would seem then to have passed north of Castell-y‑Gaer, a British earthwork with a Roman camp in the north-west corner of it, and to have followed the line of roads by Pen-yr‑heol, through Mydrim, by Caerlleon, to Post Gwynne. Fenton tells us that in 1811​41 the road was discernible in many places through the vale of Whitland after drought, and that there was appearance of a ridge near Post Gwynne; and that the peasants would track the road, called Ffordd Helen, for miles, though except where it formed the modern road there was little trace of it. Four miles west of Post Gwynne is a road in the same direction, called Park Sarnau, with a parish boundary along it for half-a‑mile, crossed by another road in a north-easterly direction, with a parish boundary along it for three miles. After an interval of six miles, a road followed by a parish boundary for four miles runs westward from Castell Hendre, past the remains of a camp a mile north-east of Ambleston, supposed by Fenton to be the station Ad vigesimum of the fictitious Itinerary of Richard. For about two miles west from Ford, and from Brawdy to Whitchurch, the road is marked Roman road on the Ordnance maps to within about three miles of St. Davids.

Parallel to, and about four miles north of this road, is the road called Ffordd Fleming — Latinized into Via Flandrensica or Flandrica. It passes along the  p374 top of Prescely mountain with a parish boundary following it for six miles, and it continues on along the tops of the hills eastwards to the borders of Carmarthenshire with the name of Hên Ffordd (old road). It slants down the south side of Foel Eryr, at the west of Prescely, and can be followed on to the north of Letterston. According to Fenton it could be traced into the promontory of St. Davids, and he correctly describes its appearance for the greater part of its length as that of a hollow way, or old unfrequented lane; though he says that on the south of Foel Eryr portions might "be distinctly traced in various stages from an open foss to the perfect raised pavement through soft ground."​42 Fenton's accuracy has been questioned, and the raised pavement cannot now be found. Many Roman coins have however been found along its course, and it is probably and older road, used, and perhaps improved in parts, by the Romans.

That is perhaps as much as can be said for the road from Carmarthen westward, but there is ground for supposing that the Romans occupied this part of the country at a comparatively early period. A hoard of 200 Roman coins found enclosed in two leaden boxes, in a camp at Llanboidy, 14 miles west of Carmarthen, in 1692, were none of them of later date than A.D. 91.43

The Itinerary of Antonine does not appear to extend further than to the borders of South Wales, to  p375 Caerleon, Usk, Abergavenny, and Kenchester, but it is evident that Geoffrey of Monmouth's road from St. David's (Menevia) to Southampton, which Higden calls Ermingº Street, and Higden's Riknild Street from St. David's to Worcester and Chesterfield, represent roads which can still be followed; though the Roman characteristics are not so plain as we go westward. The more noticeable Roman roads beyond the limits of the Itinerary are those which run across these roads from the south to the north, parallels as it were by which the mountain country was divided up and subdued.

The Author's Notes:

1 In a well at Rudge was found the curious vessel bearing the names of five Roman stations on and near the Wall of Hadrian. A Roman pavement was also discovered there in 1723.

2 Ancient Wilts, p34.

3 Itinerarium Curiosum, p142.

4 Ancient Wilts, p27 (1819). A few years later he considered that the Roman road was probably formed on a Belgic boundary bank (Roman Æra, p106).

5 Ancient Wilts, Roman Æra, p83.

6 ibid., 1819, p82.

7 Ancient Wilts, Roman Æra, 1819, p79.

8 ibid., 1812, p27.

9 ibid., 1812, p77.

10 Archaeologia, vol. II 1763.

11 Somerset Archaeol. Soc., 1876.

12 Monmouthshire, p13*.

13 Martin, Trans. Bristol and Glouces. Archaeol. Soc., vol. XXIII p309.

14 Bristol and Gloucester Archaeol. Assoc., vol. III.

15 This information from the Liber Llandavensis has been courteously furnished by Mr. James G. Wood.

16 Ormerod, Strigulensia, 1861, p20.

17 Coxe, Tour in Monmouthshire, p17*.

18 Coxe.

19 Britannia, III. 109.


Intrat et auget aquas Sabrini fluminis Osca

Praeceps; testis erit Julia Strata mihi.


21 Archaeologia, LVII.335.

22 Lysons, vol. I p200.

23 Sir R. C. Hoare, Ancient Wilts, Roman Æra, p97.

24 Itinerarium Curiosum, p67.

25 Britannia, I. 384.

26 Proceedings of the Cotteswold Field Club, vol. VI p150.

27 A belief has arisen that the use of wheeled vehicles was forbidden in the Forest, and that consequently all roads except those made in recent years must be Roman. In a summary of the "Laws and Customs of the Miners of the Forest of Dean," supposed to date from the time of Edward I, as given by Nichol (Iron-making in the Olden Time), it is stated that "carts and waynes are prohibited"; but what really was prohibited was to "make carriage" of the mine otherwise than by the measure called Bellis, so that the King might have his rights. That is, the quantity of ore was not to be measured by cart or wain load in assessing the royalty, and proves that carts and wains were in use, and that a load was then, as now, a variable quantity.

28 Trans. Bristol and Glouces. Antiq. Soc., vol. XIV p362.

29 Ormerod's Strigulensia, p3.

30 Brit. Rom., II p469.

31 Rev. H. G. Nicholl's Iron-making in the Olden Time.

32 J. G. Wood, Woolhope Field Club, 1901.

33 It was removed and in use as a garden roller for some time, and again set up on the roadside.

34 Archaeologia, IV p4.

35 Strange, Archaeol., vol. I.

36 Hoare's Giraldus Cambrensis, clvi.

37 Strange, Archaeol., vol. I.

38 Gough's Camden, III. 142.

39 Gibson's Camden, p652.

40 Hoare's Giraldus Cambrensis, p. clxiii.

41 Historic Tour in Pembrokeshire, p479.

42 Historic Tour, pp. 484, 566.

43 Llwyd in Gibson's Camden-Gough, III p135.

Thayer's Notes:

a For the contention that Silbury was not on the Roman road (with some curious consequences), see my note to chapter 1 and the further link there.

b See The Wansdyke Project 21, a comprehensive website exclusively devoted to it: maps, texts, photographs, links, opportunity to discuss it with others.

c If my edition of Ptolemy is correct on this point — a very open question, unfortunately — Codrington is wrong. Ptolemy, in the Stevenson edition, gives:

Luenti(n)um 15*45 E (of the Blessed Isles) 55°10 N
Maridunum 15*30 E 54°40 N

thus situating the former at about 55 km (35 miles) NNE of the latter (see Book II, Chapter 2 of Ptolemy and its map).

If Maridunum is supposed to be Carmarthen (modern coördinates: 4°19 W, 51°50 N), Lampeter (4°03 W, 52°05 N), today a largish town, fits almost perfectly; but Llandovery (3°43 W, 51°57 N) would be within the margins of error for ancient geography.

Maridunum was identified as Woodbury Farm, Axminster in a Report on the Geophysical Survey by M.A. Cole and N.T. Linford of English Heritage; the report has vanished offline with the continued shrinkage of the Web.

As for Llanio, Codrington's map puts it approximately where my 1:250,000 Ordnance Survey map (Routemaster 7: Wales & West Midlands) puts the village of Llanddewi Brefi; and although I've been unable to find it myself, David Petts confirms this and provides me with the Ordnance Survey grid reference, SN 6456: thank you David.

Page updated: 15 Oct 17