ROMAN ROAD Segontium to Aberglaslyn





1795   Evans J - Map of North Wales (ca 1795)

1818   Ordnance survey drawing 2in/mile in British Library cat no 306. Photocopy available at GAS.

1838   Ordnance survey 1in/mile 1st ed as reprinted Harley/Oliver, Old Series Ordnance Survey Maps, vol 6 (ISBN 0903541440)

1891   Ordnance Survey 6in/mile 1st edition 1891, as read

B        James Bransby – A description of Llanberis & the Snowdon district (1845) (Caernarfon public library)        

GAS    refers to Gwynedd Archive Service, Caernarfon

GAT   refers to Gwynedd Archaeological Trust records.

HH      Edmund Hyde Hall, A Description of Caernarvonshire (1809-1811) ed Emyr Gwynne Jones 1952 (Caernarfon library)

J         Jenkins D E, Bedd Gelert; its facts, fairies and folklore (1899) (Caernarfon public library)

M       I. D. Margary, Roman Roads in Britain, vol 2, 1st ed 1957, p82 (no 68).

R        Royal Commission on Ancient & Historical Monuments (Wales) Caernarvonshire vol ii

W       Edmund Waddelove, The Roman Roads of North Wales, 1999. ISBN 0 9506803 1 1.

WW   William Williams, Observations on the Snowdon mountains (1802) (Caernarfon public library)                                                   


M’s description, apparently derived from secondary sources rather than direct observation, is of a road approximating to the A4085 from Caernarfon through the Gwyrfai valley and then W of Beddgelert to Aberglaslyn, then the pre-turnpike road through Croesor and across to Tan Y Bwlch, and so on to Tomen y Mur.  W ch12 presents a detailed description of an alternative route approximating to the A487 passing the known but now destroyed Roman fort at Pen Llystyn, whilst his ch15 (Llanberis Pass) route echoes the start of Ms route from Segontium before diverging E to ascend Bwlch y Groes. It is conceivable that both alternatives are valid, “A4085” being earlier and “A487” being later, a situation which coincides with modern developments, since A487, although a little longer, has become the primary route whilst A4085 is the traditional route, more direct but narrow and winding, and therefore now largely confined to local and tourist traffic.


The history of the modern road is somewhat unconventional and fragmentary. Despite having been used as a long distance route for centuries – for instance Pont Peblig (ie a predecessor of the present bridge) features in a petition (see below) in 1651 (GAS XQS/1651/129) and a  “crime report” in 1652 (GAS XQS1652/66) - it is not part of the first turnpike trust in the county (Conwy-Bangor-Caernarvon-Pwllheli in 1768). Instead early improvements (eg Rhyd Ddu below) were carried out as a result of a process of  “presentment” or petition to Quarter Sessions for improving a county road, in which a number of petitioners applied for roads or bridges to be improved by the county surveyor, the cost being charged back to the parishes through which they ran. This procedure appears regularly in the surviving Quarter Sessions records, eg “Pont y Betws” (presumably Bettws Garmon) in the earliest record of 1571, (GAS XQS/1571/103), and again in 1660 (GAS XQS/1660/117). Many of these earlier references are too brief to establish whether they are of wood or stone, but by the C18th the procedure was frequent enough for “pro formas” to be inscribed first, with the details of bridge and highway added in a different hand. By later in the century we have detailed specifications for stone bridges laid out in contracts, most of the main bridges on the A4085 appearing in the 1770s (GAS XplansB/143, 147, 150, 157, 170, 214), which were generally widened from nine to eighteen feet, which they remain to this day - rather narrow by modern standards- and made more flood resistant. Although the contractors provided warranties, apparently they were not always effective because the bridge at Bettws Garmon had to be “presented” again in 1787 (GAS XQS/1787/147). In July 1796 (GAS XD2/13096) a list of subscribers, led by local gentry and contributing some £270, is set up “towards the cost of widening and repairing the public highway from the town of Caernarfon to Pont Aberglaslyn, so as to be fit for all kinds of carriages, and to open a free communication between the counties of Caernarvon and Merioneth”. By 1802 WW can say p43 “the road which is now well improved as far as Aberglaslyn bridge” (although not to everyone’s taste it seems, HH p220, referring to Beddgelert, “much amendment in the choice of a level may I think be easily had along the road coming from Bettws” – unless this is depending on secondary information, and is an out of date reference going back to pre-1777), and it was taken over by the Old Caernarvonshire turnpike trust as a result of the Act of 1810 (50 Geo III c52). Nonetheless it seems a petition to Quarter Sessions could take place even after a trust was established, eg with reference to the bridge at Beddgelert in 1814 (GAS XQS/1814/141) and elsewhere in the county (GAS XQS1795/93, XQS/1815/119 & 123).


John Evans’ map referred to hereafter as “1795” (a copy can be seen in the British library, and photocopies at GAS XM/Maps/403) is the first reasonably detailed map of Caernarvonshire, although it does not reach the same standard of surveying as subsequent OS maps. It shows the present main road much as now except for Rhyd Ddu (see below). There is more than one edition, and the Rhyd Ddu evidence shows that its compilation date could be at least in part pre-1777. In any event, it antedates much of the turnpike development and has information not present on the next comprehensive map, the OS 2in to the mile drawings of 1818.


M and modern OS regard the present A4085 from Segontium to Waunfawr as Roman because it is in straight lengths as far as Waunfawr, although this would partly parallel and be incompatible with Ws description p289 of a road running to/from Llanberis Pass The first leg which actually passes through the fort of Segontium, appears on 1818 as it is now, save that it had a distinct downstream zigzag at the crossing of Afon Seiont at Pont Peblig, a feature which has now disappeared under industrial building. However 1795 shows what appears to be an “unimproved” course at Caeathro, and again approaching Pont Crymant (ie the modern Waunfawr), which reinforces the view that the present course is the result of a late C18th improvement survey across what was at the time common land. 1795 apparently shows an unimproved course  represented now by the footpath/RoW from 5274 5917 down to the river at 5263 5905.


At this northern end however there is field evidence for a somewhat different route, although it is awkward because it starts at Bontnewydd rather than coming directly from Segontium. An RoW leaves a minor road in that village to the N of the Afon Gwyrfai at SH 4851 5997, and crosses a swampy area where a tributary flows in before turning at right angles up the hill at 4863 5983, at right angles again at 4879 5990, and a third angle at 4888 5982. By this point the path has gained flat ground at the lip of the narrow gorge through which the Gwyrfai runs for the next 5km. The second and third of these legs consist of a narrow cobbled lane, which then continues from 4888 5982 as a double field boundary, although it is not possible to walk within it because of vegetation and other obstructions. This RoW has been the subject of debate between the highway authority and landowner.


Clearly this is an old route of some kind, forming no part of modern access development, rarely used as a footpath, not appearing on 1795 and 1818, and the evidence going forward to the E strongly suggests a surveyed line rather than casual development. However the changes of course with which it starts are not necessitated by the topography, and are puzzling. A possible explanation is the presence of Plas y Bryn farm, which looks like the remnant of a small estate, although I have not to date found any history in GAS or GAT (but see RCAHMW 86461) - because it would lie across the W end of a more aligned route, and there are instances elsewhere in Britain of Roman roads being diverted when an estate was formed across them. Whilst such a hypothesis would cure the one difficulty, it would entail going down to cross the Gwyrfai unnecessarily, which the present route avoids having to do, albeit by getting involved in marshy ground instead. It is just possible that the 90 degree turns are original, designed to keep the road on drier ground for as long as possible. More research needed here, it seems.


The footpath continues on an easterly alignment, although the field evidence dies away and therefore it is constantly waymarked to show the route. 1891 marks a path but not a direct one. At 5011 5985 it closes with the modern access road and continues to Plas Glan yr Afon, 5040 5980, which is said to have C17th features (GAT12153). Just before this a side stream comes from the NE at 5033 5985, at which point the track has curved N to maintain a level course to cross it. There is some slight evidence of a more direct route going straight across the small valley on the alignment, followed by a field boundary on 1891. Beyond the farmyard it becomes a straight wide level track, maintaining an aligned course just above the main river valley, to a crossroads at 5083 5969. Beyond this it has recently been upgraded to a tarmaced road although still a track on current OS maps. At Caemabynyr 5135 5962 the made road turns N, and a RUPP (road used as a public path) is marked going straight on across the fields. Unfortunately there is absolutely no trace of, or access for, it on the ground, although there is a field boundary. Just S of the lane to Fron Goediog at 5187 5965 there may however be some slight evidence of a roadway with edging stones, if access can be obtained for a closer examination. This then closes with the lane, to which there is no access, but a field boundary is marked continuing the line at 5205 5940 towards Waunfawr. All this section of the line stays on the brow immediately above the small gorge to the S in which the Gwyfrai runs, strongly suggesting a surveyed route. It should also be noted there are a number of houses along the line, for which modern access has had to be created by running lanes down from the main road. All features appear on 1891.


At the main road at 5252 5925 the alignment is picked up by a straight narrow lane (click for photo) diverging from the main road and continuing on the NE side of the valley, whereas the present main road crosses the river and runs parallel to it on the SW. 1891 records a milestone at this junction which is not now to be found, and is not part of the turnpike sequence. The lane appears on 1818, and is consistent with the description by J pp142-3, which seems to underlie whatever sources M relies upon, “Several portions of the well-known Roman road … can still be seen…. It ran from Carnarvon through Waen Fawr, and then followed the river Gwyrfai, on its eastern (my italics) side, as far as Melin y Rhydd, (ie 547 563, where the river valley is at its narrowest) in the vale of Bettws Garmon, where it crossed over to the Western side of the river”. From Waunfawr the next 3km of the Gwyfrai runs in a wide marshy valley, very likely a filled in glacial lake, which this lane avoids, after negotiating a number of side streams, by climbing through a windgap to the E of Garreg Fawr 536 583. At 5321 5892 there is an upright stone at the roadside (click for photo) which is not obviously a gatepost, milestone?? There is a RoW branching W from the lane at 5350 5863 and going down and round the lower W side of Garreg Fawr, but there are no field traces that it might ever have been a road. Beyond 5365 5858 the lane becomes a private road leading to a farm, no public access here. There are mediaeval sites in the locality, GAT refs 4203, 5078, 7321. Beyond the farm the track passes through the waste of a slate quarry by way of an awkward turn over a small stream 5379 5821, but there is no way of telling whether this is original or a modification. It continues on a steady gradient, used as a farm track until that turns away at 5414 5804, when it becomes a green track.


At this point it becomes clear that the route as it runs now is climbing towards the ironstone mine known variously as Silurian, Ystrad or Betws Garmon 543 574 (GAT 20549), (click for photo) which has been worked like a copper mine on a lode of ore in the Nant Ffrancon series of mudstones, creating a long gash right up the hillside, very striking on the ground but barely detectable on the maps. There is little information about this mine, although surveys of parts of it were made in 1914 & 1920 (GAS X/Amp/Maps/15 & 18). It was worked in the latter half of the nineteenth century, as demonstrated by the incline from which runs down to the Welsh Highland Light railway (this stretch was built in 1877), and it is noted on 1838, although not 1818. On 1891 the workings are shown as “old quarry”, but it has been worked again since because there is a steep incline on the NW side of the lode running down towards Ystrad Isaf which does not appear on 1891; disused again by 1914. Of course intermittent working is more the rule than the exception for metal mines in this area. Many such mines in NW Wales which were formerly seen as originating early in the industrial revolution have in recent years been shown to have prehistoric origins (Great Orme at Llandudno, Parys mountain, Llanberis) and on the opposite side of Moel Eilio there appears to be a small working probably in the same ironstone vein, associated with Roman era round huts (see Roman road, Llanberis Pass section), so it is possible it had a more lengthy history.


On 1891 the track turns into a terraceway climbing on a steady gradient till it reaches the lodeline of the mine at 5433 5783, where waste been dumped across it. From this point 1891 shows a track still detectable from a distance zigzagging up the hillside and twice returning to the lode, presumably to facilitate extraction from it. Diverging slightly to the S from 5421 5797 there is a modern RoW, well waymarked, going over a stile at 5425 5781, through the lode line at 5424 5770 (beware! this whole area, like any ruined copper mine, is potentially dangerous and needs to be treated with caution. There are some modern fences although they do not appear on maps) and across boggy ground into a small defile at 5427 5742 and down to the main road at 5400 5714. As is often the case, the actual route is slightly different from the definitive map There are no discernible signs of any construction on it except a stony track S of the defile and a possible hollow way. 1818 does show a track on this course although 1838 and 1891 do not.


There is another older terraceway diverging at 5421 5797 and running to the NE of the 1891 track. It is overgrown, slightly hollowed at 5432 5788, and has then had mine waste tipped on it. It reaches the lode at 5435 5783 (photo awaited), and a grassy terrace can be seen continuing the line in the bracken on the other side. It is too dangerous to try and cross the mine at this point. Beyond this point the mountainside becomes very steep and inaccessible, particularly in summer vegetation, and no traces can be observed from a distance. The stone wall running E up the mountain utilizes a prominent little outcrop at 5439 5779, which would be slightly above the line, and makes a useful landmark. S of the wall there is a shallow regular excavation running roughly parallel with it, eg at 5436 5776, which looks artificial, might this be a continuation of the possible boundary ditch on the opposite side of Moel Eilio at 5600 5965 (GAT5024)?


Some 400m further SE, across the lower end of the scree slope below Diffws on approximately the same 240m contour, there is what on superficial examination appears to be a sheeppath running from 5457 5743 through 5459 5738, 5462 5733, 5465 5728, 5470 5717, and across the crags to 5474 5710. It can clearly be seen, even in summer vegetation conditions, from the main road below. It runs virtually in a straight line, initially gently downhill, from 5462 5733 uphill again, and appears to be the last remaining trace of a terraceway, whether filled in or washed out is not immediately clear. There is so much loose stone that it is difficult to make out any real signs of construction, and it is not marked on any map, neither are there any modern paths in the vicinity. This line is precisely where a continuation of the road would have to be to work its way through this very broken ground. It is walkable without any real difficulty except at the S end where it makes its way across a near vertical crag. Once the top of the crag is reached, the only way forward is to drop down some 120m through a very steep side valley directly on to Melin y Rhydd. However, this valley is waist high in bracken in summer conditions, when nothing can be seen on the ground. A winter survey might yield some evidence; one might search for zigzags, since a direct route would be almost impossible even with the Roman military tolerance for steep ascents. If it is of such an origin, it is easy to see how difficulty and erosion would have led to its early abandonment.


At first sight this route from Waunfawr to Melin y Rhydd seems improbable. From Waunfawr E to the mine it has evidently been a mine access road, but the question is whether the line of access has an older history. There are five items of evidence which suggest it has; i) 1818 shows it with no mine, ii) Js remark quoted above, since no other course seems possible E of the river, iii) the apparent continuation of the alignment W of the main road from Waunfawr, which makes no sense in the modern landscape, iv) a number of archaeological sites lie along it, eg the mediaeval township of Treflan 536 586 (GAT 7321 et al already noted) and beyond the mine GAT 4200, 4210, 6717, 6745, 6746, 6748), v) the disused terraceway disregarded by the mine workings at 5435 5783. In comparison with the present main road, it seems rather a detour, the route has climbed high above the valley - by the time the ironstone mine is reached sufficient height has been gained to enable line of sight with Segontium - only to come down to river level again through very difficult terrain, but on the positive side the route avoids the valley and a large expanse of flat marshy ground below, which would have been a lake at some time in the past. It is the climb up the hill that makes it seem a detour to modern eyes, not the additional distance, which is insignificant.


The main Gwyrfai valley now runs SE as a narrow trough exiting from Llyn Cwellyn, with a “pinch point” between crags at 547 563, the Melin y Rhydd of Js description, any route could only have passed through here. When the valley opens out again, on the NE side of the river and lake there is only the most fragmentary evidence - a possible hollow way at 5665 5522 and the slightest hint of a terraceway at 5740 5459. The track marked on modern OS 1:25000 at 5740 5485 etc, which forms quite a prominent feature on this hillside, is a modern machine driven enterprise.


However, tradition places the Roman road firmly on the SW side of the lake. J p143 (continuing from above)  “It then passed right underneath Castell Cidwm, climbed up Gallt y Llyn (the slope above Cawellyn), where it can still be seen”, and 50 years earlier Bs description p52  “vestiges of a Roman road called Ffordd Pawl…the celebrated road which passed by the edge of Cwellyn Lake”. M doubts the possibility of this because he seems to think there is no room between lake and mountain at the northern end, but this is not the case, there is a terrace/RoW on this side, which could answer to it. At the pinch point there is a bridge 3m wide of indeterminate age crossing Afon Gwyrfai, Pont Cerig y Rhyd 5465 5634, which is shown on the undated engraving of the now demolished Nant Mill (GAT64923) GAS XS/3092/2. This now carries a track and RoW, the RoW following the track until it turns W 5459 5620 up the cwm to a dead end. The track, prominently signed as “private no access”, continues S to Cwm Bychan farm 545 560, but there is nothing beyond that in the fields until Castell Cidwm is reached. In its present form the northern end of the track is blocked at Castell Cidwm slate quarry 552 552 (GAT 20185), although 1891 shows that track curving past the end of the lake (and therefore over the river) to the main road. To the S the footpath of 1891 and modern RoW have been expanded into a modern forestry road, with no trace of anything older, but where it drops down to the camp site at the S end of the lake 5657 5408 a slightly hollow lane can be found along the wood edge continuing the alignment to a ruined building 5673 5382 and out into an overgrown meadow where traces of a terraceway continue at 5677 5376, most clearly observable from the main road below. Beyond another ruined building a stone pillar for a gatepost at 5702 5342 would mark the same line, then the waste from Cwellyn slate quarry precludes any further evidence. This lane is the only clear remaining evidence of an old road line on this side of the valley. The line would bring the older road back to the A4085 at the T-junction just above Rhyd Ddu bridge 569 529.


But J already quoted states the Roman road “climbed up Gallt y Llyn”, which implies we should be looking at a higher route represented by the bridlepath (definitive map ref 38) shown as RoW on modern maps climbing to Bwlch y Moch, there is a modern forestry road roughly corresponding to the line of this path which crosses this very extremely steep slope as a terraceway, although it does not appear on any OS maps at all until RoW information is added in 1971. The definitive map does mark a very slightly different course from the forestry road, although there is no evidence for it on the ground. The last few metres of the track has been obstructed by fallen trees until the exit from the forest is reached at 5608 5401, where a wicket gate originally allowed passage through the stone wall (photo awaited).  Crossways there is another RoW coming up from Planwydd by the camp site 567 540, used by modern walkers to access Craig y Bera 543 542, which in 1891 came through the same gate, but has been relocated by Forest Enterprise onto the line of former electricity pylons and now crosses the same stone wall 20m above the pass by a ladder stile at 5605 5408. 1795 shows this route as a road running down into the Nantlle valley, ie the predecessor to the 1810 turnpike/modern B4418, and R p lxiv considers it to be a packhorse trail. Felling is taking place in this forest in autumn 2003, so more evidence may be observed once it is cleared.


Bwlch y Moch gives a good view of much of the surrounding country; in addition to J’s comments, the route is more direct when “1795” is seen as the southward continuation rather than A4085, and it gives easy access to Drws y Coed copper mines if they were active (possible but unproven) when it was devised. The hypothesis that this could be the original course seems therefore to be a reasonable one, but the lack of any confirmatory field evidence is frustrating. If correct, again we have a road that chooses high ground, not valley bottom, as its preferred route.


Beyond the pass the bridleroad is mapped coming to an abrupt end close to the shore of Llyn y Dywarchen, but that does not appear on 1891 and it cannot now be traced on the ground. 1891 does show a footpath going slightly E of S, through the now abandoned farm of Llyn y forwyn past the E end of Lyn y Dywarchen, whose size grew considerably when it was converted into a reservoir for Drws y Coed copper mine in the nineteenth century. The 1891 path came down to the present road at 5627 5290. No trace of this path now remains in the peaty ground and the S end of it is steep and awkward. However, later OS 6” surveys show an alternative path a little to the E, and this is reflected by gates in the walls at 5641 5277 and again at an oblique angle to the present B road at 5641 5278. The Sward continuation from the latter point was in 1891 the only access to the house now called Drws y Coed Uchaf 5642 5266, and such a route would give a connection straight into the “1795” route described below. It has only been during the C20th that the farm name migrated to this site, originally it belonged to the building by the dam on the B road at 5590 5332, where the name makes better sense as a pair to Drws y Coed Isaf in the Nantlle valley, and this is the location to which R no 712 refers.


From Rhyd Ddu the modern road is as now from 1818 on, straight to the top of the shallow pass at Pitt’s Head, with the trackbed (soon to be reconstructed) of the Welsh Highland Railway taking the same course first on E and then to W. For the origin of the name Pitt’s Head see J p152 – it does not refer to the top of the pass. Then the A4085 crosses Afon Colwyn by Pont Caer Gors, 5758 5091, but 1838 and 1891 show a by road leaving here and running on the W side of the river parallel with the present main road. Although marked only as a RoW on modern maps, this is now a modern forestry road leading to a car park, and then a drivable track running down past Hafod Ruffydd Isaf to its own crossing of the Colwyn at Pont Hafod Ryffydd or Pont Glan y Gors (1891) 5748 4997, rejoining the main road as a tarmac lane a little further down. The main road is a slightly longer and gentler ascent, and does not appear on 1818, demonstrating that it is a “turnpike mk2” and the by road “turnpike mk1”, just like the A498 coming up to Pen y Gwryd from the S. R lxvi states this road was under construction between 1776-8 and followed a previous road except for “deviations between Rhyd Ddu and Lion Rock passing West of Llyn y Gader”, except that the authors must mean E, not W. The source of this information is not stated explicitly, but is verified by GAS Xplans/B153, which is a contract with the Quarter Sessions clerk (see above) dated 3rd October 1778 to build “a new bridge over Afon y Gader at Y Rhyd Ddu [this is a substantial stone skew bridge on the main road] on the new road leading from Quellyn bridge to Pont ar Golwyn”. The latter must be what 1891 calls “Pont Hafod Ryffydd or Pont Glan y Gors”, since 1818 shows the only other bridge over the Colwyn, Pont Caer Gors, did not exist at the time. The 1778 contract for Golwyn, GAS Xplans/B143, is to repair the existing bridge, not make a new one, and similarly for Quellyn GAS Xplans/B150, showing that the other parts of the present road antedate the modernization. A Pont ar Golwyn, presumably the same, is referred to in 1660 (GAS XQS/1640/M2); Beddgelert bridge itself is presumably that referred to in 1657, (GAS XQS/1657/199)


1795 shows a more circuitous winding course going W of Llyn y Gader. This is about ¾ mile longer than the present road, nonetheless the mileage figures from Caernarfon on some editions of 1795 are located at the current turnpike milestones, ie they reflect the shorter route not the longer one, presumably because the survey was done prior to 1778 and the mileages added later. This route is supported by J p143 “[Roman road] passed west of Llyn y Gadair, over the hill called Y Gadair, to Glan y Gors.” Just above the bridge the B4418 (itself turnpiked around 1810) branches from the A4085, then at 5668 5263 the modern road turns sharp right, and 1891 records a “tablet” at this point with mileages to Llanfihangel and Dolbenmaen – this refers to the “bridlepath” that goes on up the hillside, through what is now forest, and ultimately into Cwm Pennant by way of Bwlch y Ddwy Elor (“dark and magnificent passage” HH p232, GAT 9648-9659). This pass is generally regarded as an “old” route (R p lxiv again) although there is no evidence of construction going up the hill on this side, it is a footpath rather than a proper bridleway. It does appear to be the route described by GAS XQS/1735/41 & 42 “the bridge of pont Llanfihangel on the road between pwllhely and Llanrwst was in 1726 presented by the Grand Jury at the Quarter Sessions to be ruinous and out of repair”, and mentioned even earlier GAS XQS/1657/184 and XQS/1658/E/22. The “tablet” must be that which is now set into the roadside wall back at the road junction in Rhyd Ddu, it is carved out of cut slate and is of the same type as those at Aberglaslyn bridge, I guess of C18th/19th origin.


From 5668 5263 this bridleway continues the line of the road as a modest causeway (road on 1838 as far as the stream) going straight on from the junction across gently falling ground to a small stream at 5645 5246, which it now crosses by a vehicle-width platform of old railway sleepers, then climbing the hillside as a footpath. Nb hut circles at 5622 5212 GAT597 “Roman”. Our line is nominally represented on modern OS by another bridleway almost at right angles crossing a little further up the hill to the W. Presumably this is the route referred to by Bp52 “on the S side of Drws y Coed may be distinctly seen the traces of a still more ancient road, Mignedd, which forms a passage…to Cwm Marchnad”; and also the road marked on 1795 as connecting at Drws y Coed with the packhorse trail from Bwlch y Moch already referred, by ascending gently to the present B road at Bwlchgylfin. However it is not shown on 1818/1838/1891 and there are no obvious traces on the very marshy ground. 1891 shows a footpath starting from the head of the quarry tramway across the marsh by the lake, falling in with the modern RoW beyond Gader Wyllt at 5648 5180, the second of the two small quarries. However, there are traces of two other routes across this ground, the first through a galvanized iron gate at 5642 5237, from which a terraceway can be traced leading up to the first quarry Llyn y gader – this presumably was the access to it before the tramway was built further E. The second less distinct route is represented by a blocked Victorian era wicket gate at 5635 5237, then a galvanized iron gate at 5634 5224 followed by more railway sleepers over a stream, and a third set of sleepers across a watercourse at 5630 5222. At the next stream there are no bridge remains, but a slight terraceway climbing the S side at 5632 5210, followed by slight hollows or terraces at 5637 5193, 5642 5188, etc., although there has probably been some disruption hereabouts from slate waste tipping. A modern ladder stile at 5642 5178 is the first evidence of the actual map RoW, then a terrace 5644 5178 climbs away from the stream, apparently forming an upstream zigzag to get over the watercourse before turning S at 5648 5180 on the line of a stone wall. This whole area is a shallow peaty hillside with impervious slate near the surface, and little or no boulder clay to soak up any water, hence the many streams.


The stone wall now runs almost due S for several hundred metres, with evidence of a road variously underneath or to one side or the other, evidently laid out using the road as a guide but after it had gone out of use. At 5652 5178 there is a prominent sinuous hollow way up to 2m deep on the E side of the wall which eases the climb up the hill, with a much slighter hollow immediately under the wall itself. A terraceway can be seen at 5654 5167 forming the leg of another upstream zigzag to a watercourse which has recently had the peat removed to improve drainage, and in the bottom at a depth of about 1m there are clear traces of metalling 5655 5163 (click for photo). The opposite leg is a slight hollow 5654 5164, 1891 clearly showing this zigzag as the course of the footpath. More peaty ground follows with no distinguishing features. 


At 5655 5144 the RoW passes into Beddgelert Forest. A slightly more direct course a little higher up Moelfryn was taken by the stone wall, which may be worth checking if it can be found in the undergrowth at all. This is one of the older forestry commission sites, and much of this area has been ploughed, cleared, and in places allowed to regrow naturally, so that other than the access roads it is often impenetrable, and any field evidence almost impossible to detect; GPS readings can also prove unobtainable or unreliable. Our path seems to run in the bottom of a shallow hollow way until it joins a forestry road, 5655 5128, (check) waypoint 47 in the Forestry Commission pamphlet of forest walks obtainable from the campsite at 578 490. Walking in the opposite direction this path can be located immediately by the waypoint. The forestry road is the modernized version of one shown on 1838, in which it appears to be an access road coming down from the old level and small quarry on Bwlch Ddwy Elor GAT 9648, which was by the packhorse trail to Cwm Pennant referred to earlier. From waypoint 47 the forestry track continues the alignment of the old road SE to waypoint 63, 5697 5079, where 1838 turns E to reach the main road at Pont Caer Gors


This whole section from Rhyd Ddu would have gone completely out of use once the new road was finished in 1778, and high rainfall and peat have rapidly obscured the original course.


Picking up from waypoint 63, on initial examination three possibilities for a line can be detected on 1891.

                                         1. On 1891 there is a footpath/wall (just wall 1838) continuing the alignment from waypoint 63 and then curving more southerly to reach Glan y Gors at 5731 5013, a site now lost near the present forest car park at 5735 5026 - J p143 above talks of the Roman road going to this farmhouse, but no trace beyond. This section is now impenetrable undergrowth. The present access road to Hafod Ruffydd Ganol/Isaf crosses just below the confluence of Afon Hafod Rhuffydd Isaf and Afon Cwm Du, by an arched bridge Pont Rhyd Cefyllau 5732 5001 (GAT9674, waypoint 87). There is a modern path from 5735 5026 down to the bridge, with some suggestion of a hollow way in the undergrowth immediately to the W, check in winter conditions, but this would be a bit offline The bridge appears on both 1838 and 1891, and on its northern side there appear to be two intact dry stone abutments, large fairly regular blocks, capable of taking a timber predecessor (click for photo). This is unusual and probably significant. There is also a ford just S of the bridge, although this has obviously been used in modern times by forestry vehicles, since the bridge itself now has a 5 ton weight limit. The track proceeding from the bridge is not straight, but skirts marshy looking ground and crosses the hillside below Hafod Ruffydd Isaf (Beddgelert Mountain Bikes), now constructed as a substantial forestry road which approximates to the 1891 track but inevitably has no historic traces.

                                         2. 1891 also has a footpath starting at 5697 5075 (a lost building a little to the S of waypoint 63) through footbridges 5702 5025 and 5703 5015 (these crossings being very close to what is now the trackbed of the Welsh Mountain railway, built after 1891) and then closing with 1 near Hafod Ruffydd Uchaf (see below) at 5707 4982. Other than a gateway at this last point, I cannot find any surviving traces on the ground.

                                         3. A footpath W of Glan y Gors on 1891 utilizing stepping stones 5722 5012 (GAT9672) and footbridge 5719 5007 (GAT9671) before falling in with 1. No surviving traces.

2 is the most direct route, but the disadvantage of both 2 and 3 is that the area they cross is riddled with small streams issuing from a spring line just below 1 on the SW, which must have made the area very marshy and awkward, whereas 1 stays on higher if slightly less direct ground for as long as possible before coming down to a single river crossing, rather than making multiple crossings. At Glan y Gors the post-Roman track would have gone straight to Pont Hafod Ryffydd/Pont Glan y Gors/Pont Golwyn to go on down to Beddgelert as it does now. 


M suggests a route above and W of Beddgelert past the farmhouse of Meillionen, and this follows J p143  - continuing from last quote - “We have no trace of it again until we pass Meillionen, and then it comes to view above Ty’n y Coed, whence it can be followed until within a field of Cwm Cloch”. The terrain allows this as a viable route for a road, and its likely age is suggested by the fact it completely ignores the village/river junction of Beddgelert, whose development originates from a religious foundation generally considered to be of C6th date. The course is roughly represented by a RoW marked on 1:25000 as a bridleway, but it is quite unusable as such, indeed much of it is unusable even as a footpath. The first part of the map route crosses Afon Cwm Du by a footbridge 5745 4991 (“stepping stones” 1891) and is waymarked for a few metres before becoming untraceable on the cleared forestry ground. At 5716 4923 an easily missed path bears to the right, (waypoint 5 on the forestry map but the actual marker is missing) and leads down as an overgrown track to cross Afon Meillionen by way of a substantial bridge at 5721 4919 (click for photo); this is a single arch entirely of dry stone construction some 3m wide. It appears to have been built within two abutments some 25m apart, again suggesting the possibility of an earlier wooden bridge albeit a rather long span. Like Pont Rhyd Cefyllau it is carefully sited just below a confluence but before the valley becomes significantly incised. The location is obscure, J does not mention it and despite its obvious age it has not been registered on the Sites and Monuments Record of GAT, and I can find no name for it. The ford recorded further up Afon Meillionen GAT9676 5686 4874, waypoint 7, is on a path/RoW connecting Beddgelert with Cwm Pennant via Bwlch Meillionen, and is some way off course.


From the bridge the RoW is marked going SE through the farm of Meillionen but there is no trace on the ground. However, there is a slight terrace rising from the bridge a few metres to the W, a dump for tree branches in 2003, and 1838 & 1891 then mark a path parallel to the present nominal RoW running into and through a double field boundary above the farm at 5739 4872. The whole stretch from Pont Rhyd Cefyllau to this point is shown on 1838 as an approach to Meillionen itself, a farm of at least C17th origin. Through a stone wall, in which there is now no break or sign of reworking, 1891 then has a path running through Parc Ty’n y Coed, although neither it nor the “bridleway” further down can now be traced until the edge of the forest at a stream 5775 4848, waypoint 9, ford on 1891. Out on the rough pasture the path can just be traced going S – there is little constructional evidence but at 5799 4806 water running across the line has exposed a metalled surface with a slight terraceway coming down from the N (click for photo). This stretch coincides with J’s description p143 above, and more specifically at p130 “ [Mr. Thomas the farmer at Cwm Cloch Uchaf] very kindly took us along a lengthy portion of the old Roman bridle path, which can still be unmistakably traced through three or four of his fields”. There are indications the present landowner is rather less accommodating to walkers. Beyond Cwm Cloch Uchaf farm there is a track continuing the line, although it seems not to be present on 1891. At 5837 4760 the RoW enters the grounds of Bron Hebog and according to notices at the gate there in 2002 has been extinguished.


J p143 [following on from previous quote] “It can scarcely be doubted that the portion obliterated made straight for the front of the place where now stands the Royal Goat Hotel. From here it followed the same course as the present road, until near the old engine-house on the right hand side of the road. It then crossed over Bryn y Felin, passing through a cutting in the rock by a little tower, (mentioned again at p310) and crossing over the very mouth of an old copper level, to emerge once more into the main road, about ten yards south of the third telegraph post from the old stamping houses on the roadside. We know nothing definite of its course again until we are right in the pass. Some think that parts of it can be traced on the rocks above the road,” then p317 “near the bottom of the hill, between the wall and the river, is a cluster of fir trees … thriving on a small embankment, which projects towards the river at an angle of something like forty five degrees. The embankment is a part of the old Roman road, and it cannot be much more than twenty years since one of the abutments of the bridge was washed away…Just opposite this little grove lay the pool called Llyn yr Hen Bont… which was made possible by the strong weir which existed halfway between the Roman bridge and the present bridge”.  Aberglaslyn old bridge is recorded at 5953 4642 (GAT6644)


J p310 “Some five yards beyond the Aberglaslyn Bridge, on the Nanmor side, we see an old cutting above the road, on our left, which marks out its further course.” 


Llwbr Rhys Goch


p333-5 “ancient little building is Pen y Groes – so called from the fact that the old road branched into two at this point” p334 “A few yards below Pen y Groes one may see traces of an old road, twisting round so as to cross the present road leading to Penrhyn Deudraeth. It skirted Bryn Gelli’r Ynn … and at the farthest end of the hillock crossed a little bridge, which some think to be of Roman origin. By stooping to peep under the arch, it is easy to see that the little bridge did service for generations as a narrow bridge of a bridle path; and its width was, doubtless, doubled to cope with the increased business of the quay. It spans what was once the bed of Cwmcaeth rivulet….A causeway once ran along the marshes as far as the bend of the river opposite Dinas Ddu; then a winding path mounted the bend in Bryn y Gwynt, where it turned to the right. This path must have branched off in two directions near the lake of Hafod y Llyn, the one making for Hafod Garegog…and the other passing Hafod y Llyn and Ferlas. The former is called Llwybr Rhys Goch…It is excellently constructed, resembling even in details the recognized Roman roads. In order to guide the foot over the swamps, should it be dark, stones were set on edges on each side. The path led on by Ynysfor and Careg Hylldrem, turning off at the latter place to the left, as the tide washed right up to the bridge… Many take this to be a genuine Roman road, made to join Aberglaslyn and a road in Lower Merioneth, somewhere near Penrhyn Deudraeth”. p336 “Hafod y Llyn Isaf is at the lower end of Bryn y Gwynt, and a path from there to Ferlas is kept open by a beam over Nanmor river, where the ruins of an old stone bridge may be seen, stretched across the river bed”.



RCAHMW 23504 “old trackway” Llwbr Rhys Goch 5985 4514? And 6040 4435 GAT 48 “Roman”


Copyright John Byde.    Revision date 30 December 2003