This blog contains research about Roman roads and other old routes in northern Gwynedd, where I now live. I first learnt to read an Ordnance Survey map when I was about 6. Although I didn’t know it at the time, I crossed Margary RR14 at TQ35767332 every day in later primary and all of secondary school, to and from my grandmother’s home nearby. When I was 17 I worked as a volunteer at Lullingstone Roman Villa in Kent, so learnt some basic archaeology there. I met Ivan Margary once on-site, and started taking an interest in Roman roads, and communications generally, round about then. I bought the first edition of his two volume work, and must have read it pretty thoroughly over the years. My degree is in Latin, Greek, and Theology, including researching Byzantine liturgy, and my working life has been spent largely in childcare. Over the last 60 years I have walked many roads, Ackling Dyke in Dorset, Dere Street over the Cheviots, the service road for Hadrian’s Wall in some places, and lots of others. I’ve seen the Roman bridge at Piercebridge. But I am not a professional archaeologist, so I don’t have access to modern resources and techniques. What I am covering in the following pages is old-fashioned fieldwork, with some historical research and a smattering of modern technology. More to follow later.

Ashton’s road to Parys Mountain

The road Margary RR67 is attested as Iter XI of the Antonine Itinerary, which is the shortest entry in terms of the number of stages. Segontium is a well attested, excavated Roman fort and vicus, which has an afterlife in several Welsh legends. It is reasonable to think of it as the most important Roman presence in North West Wales. It was probably founded by the Roman governor Agricola in AD78 following his victory against the Ordovices. The tribal name survives now as Dinas Dinorwig, and thence as the village name which has migrated a few miles uphill. But this was not the first Roman presence here, since in AD61 the governor Gaius Suetonius Paulinus engaged with the “Druids” in a battle on the Menai Strait (Tacitus Annals 14.29). He then learnt of the Boudiccan revolt started in Camulodunum (Colchester), spreading to London and St. Albans. He had to return south in order to put it down.

The Roman navy would not have been capable of conveying messengers and troops (even allowing for Tacitus’ figure of 10000 as an exaggeration) around the south western end of Cornwall in such an emergency. There must have been a route linking the Menai Strait and London in AD61, but where did that battle take place? I think we are in danger of being blinded by modern topography, most historians assume it must have been either Bangor Ferry, or Moel y Don. But it is well attested both by literature and maps that the main crossing up until the inundation of Lavan Sands, was at what is now called Beaumaris. That is why Edward I built a castle there, to dominate the crossing, just as at Conwy and Rhuddlan. The road over Bwlch y Deufaen can best be viewed as a Romanized trackway, which continued across the Menai Strait to Parys mountain, as described by William Ashton. The section running from Gorddinog to Segontium was added some years later, when the latter fort/town was created in the latter half of the 1st century.

Extract from William Ashton, The Evolution of a Coastline (1920)

Mr. Elias Owen writes about this road : — ” I well remember 35 years ago (circ. 1859) noticing the ridge or raised roadway then existing, and which there was every reason to believe was a Roman road, raised and paved, going over the sands, and which most likely ended in a ferry opposite Beaumaris.” An 1805 map (Laurie & Whittle, London) shows two distinct tracks to Beaumaris, the one commencing from between Penmaenmawr and Llanfairfechan. and the other from opposite Aber. [1795 shows the same configuration, with the more northerly line marked as a coach road]

Early in the 19th century the fine new coach road through Nantfrancon and the Llugwy Valley, by Corwen, Bettws-y-coed, and Capel Curig, diverted the Irish traffic to the Bangor ferry, the Straits being crossed near the spot selected by Telford, 20 years later, for the building of his great experimental Suspension Bridge. This bridge was opened for road traffic in 1826. Four years later the railway era was inaugurated, and in 1850 Stephenson’s great Britannia Tubular Bridge, about two miles west of the Suspension Bridge, diverted to the Chester and Holyhead railway the bulk of the Irish traffic which Telford’s bridge had been designed to carry. The way from Conway to Beaumaris was joined, before reaching the ferry opposite Beaumaris, by an ancient track which came from the Roman station of Kanovium, or Caerhun (five miles above Conway, up the river Conway), over the mountains through the high pass of Bwlch-y-ddeufaen, and descending to the coast through a narrow valley, Rhiwiau, on the Aber side of Llanfairfechan, opposite Bryn-y-neuadd. In 1883 two Roman milestones, of dates between 119 and 211 A.D. [see GAT4066] were found in this valley, proving by the inscriptions upon them that this road had been used by the Romans as a way of approach from Kanovium to the copper mines at Pary’s Mountain. This Kanovium track over the mountains is one of the earliest of roads across Carnarvonshire of which we have any knowledge. It was part of the Roman road from Chester through Denbigh, Llansannan, and Llan-gerniew to Segontium ; others think through St. Asaph and Bettws- Abergele ; the evidences are very vague. Near Careg Vawr (overlooking Llanfairfechan) the road forked, the left leading through Aber to Segontium, and the right towards Anglesey. The writer has taken some pains to trace the way on Anglesey taken by the Romans to reach the Amlwch copper mines. Precisely opposite the old road across the mountains from Kanovium, near Llanfairfechan, about 3i miles east of Beaumeiris on the Anglesey coast, the old way probably went between Penmon School and the sea ; east of Cornelyn, Llangoed ; above the Methodist Chapel and past the windmill ; then, after a sharp turn to the left, to the hill Marian Dyrys ; by Coed Cywydd Farm and Dinas Sylwy, and thence direct to Red Wharf Bay at a point nearly a mile below its head. Here it is joined by a direct road from Beaumaris, which would be an alternative way from across the Straits. Subsidence having caused the sea to advance up Red Wharf Bay to this extent, the old track is now submerged for about 21 miles of its course. On the opposite, or west, side of the Bay a well-defined road ascends from close to the north side of the Ship Inn ; thence by Llanbedrgoch and Penrhos Llugwy (where quantities of Roman pottery and coins have been found) to the Copper Mines at Pary’s Mountain. It seemed worth while to make this examination, as the support which it gives to the subsidence theory is incontrovertible. A marked characteristic of Roman roads is their directness, a condition well fulfilled in this case.

” The remains of a paved Roman way, according to ” Lewis, may be traced leading through Penmon, ** towards Llaniestyn.” (Williams’ ” Guide to the Menai Straits.”).

This submerged road, whose existence is evident to anyone looking at a detailed map of eastern Anglesey, must be reflect a time when sea level was considerably than it is now, and Penysarn (top of causeway) at the northern end SH459904 is a significant name. Beware of proceeding too far down to the inlets by vehicle on this road, it is very difficult to turn! Now we know that Parys copper mine was worked in the Bronze Age as well as by the Romans, it fits very logically as an ancient trackway like those running east/west further south across the Rhinogau.


1795 Evans J – Map of North Wales (ca 1795). Available online from National Library of Wales.

William Ashton, The Evolution of a Coastline (1920)

GAS refers to Gwynedd Archive Service, Caernarfon

GAT refers to Gwynedd Archaeological trust records.

John Byde 7th April 2022

Margary RR67 (part) Canovium (Caerhun) to Segontium (Caernarfon)

For a road that is listed in the Antonine Itinerary parts of this route remain elusive, despite the abundance of milestones associated with it. Most of it coming up from Canovium/Caerhun to Bwlch Ddeufaen is well established, and a milestone was found in 1954 in the vicinity of SH71987155, GAT4688. This relatively high level pass (1375ft=419m) is historically the safest way of reaching north West Wales, since it avoids the difficulties of travelling down the coast from Conwy (itself founded by Edward I with the building of the castle) past Penmaenmawr and Penmaenbach, where there is a history of unwary travellers falling in to the sea. The prehistoric standing stones GAT523 suggest the Romans were not the first to use it as a long distance route. See my separate article Ashton’s Road to Parys Mountain. The same course is used by the National Grid powerlines established in the 1930s, and the wide well established track that is generally referred to as the Roman road on OS maps was utilized with some modifications to enable their construction. The route was also included in a turnpike Act of 1777, and possibly some work undertaken.

Travelling west from the top of the pass there are indications where the older route is visible to one side or the other of modern track. By a ford SH70307207 there is an upstream zigzag, and W of this the earlier course is marked by a hollow on N side of present track. After a second ford at SH69927222 the modern route bears to the south uphill, crossing the shoulder of Yr Orsedd SH69307224. However 1891 marks a separate track to N, following the contour in an arc above the more direct stone wall up to the crest at SH69207251. All that is visible on the ground is a sheep track, but this might reflect the presence of drier ground at the outer edge of a filled in terrace. It is no longer marked as a path.

The section in italics I walked and described before the road was proved along this route by excavation at SH69437236 (GAT 217571), SH69437228 (GAT 17572), SH67807257 (GAT17841). The crest seems significant because it an excellent aiming point from Gorddinog below, a clear saddle in the profile of the spur of the mountain. The three milestones found on or near the Gorddinog route, two at GAT4066 SH67907295 (there is a replica in the field of one of them), 2nd and 3rd centuries, and at Madryn SH66887338 GAT368 (“possible milestone”), plus its very clear military advantage going down a spur of the mountain with extensive views, seems conclusive enough.

At this point Margary’s text seems confused, because he refers to a narrow terrace zigzagging down to Gorddinog. His source, it is unlikely he would have walked it himself, is Hemp, and that text, which describes what was known of the road at the time, but in the reverse order, clearly states the route via Gorddinog because of the milestones GAT4066, which were discovered in 1883, but it doesn’t mention the zig zags. I suspect what Margery saw were the zigzags above the end of the tarmac lane SH67607160, where there is now a small parking area. Even now some OS maps mark them as the course of the Roman road. Such a series of zigzags is a familiar feature of packhorse trails, eg above Beddgelert SH58644709 and elsewhere, and these are narrow and slightly hollowed, with no evidence of a constructed road. This line is littered with cairns, enclosures etc of uncertain date, so the route may even have existed before the Roman period. The tarmac lane which goes down via Bont Newydd and on to Aber is referred to in 1652 GAS XQS1652/156, where it is called “hwylfa gogvfryn” (? halfway valley)

Returning to the direct route, there is a track continuing the line down from the spur although it curves away N at SH68857272. The straight on course is very steep although it does show from below as a grass sward in the heather, so there might be metalling underneath. Either that, or a small zigzag as shown on 1891. A wall comes onto line at SH68797269, and the route is marked as a footpath on 1891 and right of way on current maps , and as a road throughout on 1838, although noone seems to walk on it now, due no doubt to its inaccessibility further down. From SH68357285 on it develops as a very wide deep hollow way into which rocks to clear the surface for pasture have been dumped, petering out at SH67967309. The right of way to S at this point, which does have a stile, goes through a wood to Rhiwiau Isaf, but this is now a riding school and the wood is used as stabling for a large number of horses, which makes it very wet, slippery, and potentially dangerous. The wall goes straight on to a T junction at SH67867330, although the ground is now featureless. Beyond the junction there is a right of way running at right angles, but there is no path and it is blocked at SH67757327 by a barbed wire fence. Below the fence however a wide slightly hollowed terraceway continues down the hill turning slightly N of W, and is accessible from the by road below by a stile at SH67607322. This runs out into the field where there are no visible traces.

Once the low lying land is reached, there are no obvious traces at first. The terraceway at Pen y Bryn SH65707283 GAT11168, is considered to be an early version of the turnpike.

The next evidence is somewhat to SE of this alignment, approaching the farm Henffordd SH65227232. There is a footpath leaving the byroad to Bont Newydd (which itself is the continuation of the later diversion described above) at SH65787248 (not the track going straight up the shoulder of the hill) and curving round the hillside to the farm. At SH65527255 it is joined by another path coming up from the village past Tan yr Allt, which would be a better link to the terraceway on the other side of the river, although there is no field evidence as such on it. From this point on there is a succession of evidence running SW, and seen from vantage points on the A55 below the guiding principle seems to be that the line runs in the angle between the mountainside and the coastal plain, which is often the break point between cultivation and rough pasture. There are a number of farms on this line, whose only access now is by individual tracks from the N, which could be of significance.

Slightly further down the plain is the byroad from Aber church, which runs to Crymlyn (marked here by 1891 rather implausibly as Roman Road, considering it is formed of a series of right angled bends), then below the former Penrhyn Estate stables at Tyn yr Hendre, St. Cross church, down a hollow way at SH60657085 which was truncated by the railway in 1849 (by which time the turnpike would of course have superseded it) and across the Ogwen at Tal y Bont. There is then a by road/footpath recrossing the railway to the upper part of Llandygai, and a footpath between the railway and Industrial Estate to a crossing of the Cegin at SH58827115, with a hollow way continuing up the hill, then a minor road on Maesgeirchen estate and a metalled track over Bangor mountain (which W, I think mistakenly, considers to be Roman) and down to the city. Presumably this is the pre turnpike route again. The by road and ford at the Cegin at SH58677065 could be a variation of this – fords still in use are very rare in this area now.

SW of Henffordd there is a line of very large oak trees, the footpath runs on the S side of them with a terraceway (although it looks to have been modified by modern use) then a field gate at SH64807210. If approaching from the opposite direction, there is a wicket gate slightly to the N of this, don’t use it because although it might seem to be the footpath there is no (legitimate) way out of the succeeding fields. Nothing evident behind Glyn farm, there is a footpath running SW of the building and on to Plas Nant which is featureless in itself, but presumably is a memory trace of the road, because the field boundary at SH64557180 has evidence of a ruined terraceway.

Nothing clear for next half mile (unless the boundary of Crymlyn Oaks wood has evidence, would need landowner permission; similarly in fields between Crymlyn and Gilfach). Then at SH63207142 there is a field boundary, with wicket gates and footpath on N, lower side. However, an earlier edition of OS shows the footpath on the southern side (but 1891 shows nothing), and there is a clear terraceway running parallel with the boundary here. After a break in the next field it is again seen W of Tan yr Allt at SH63907129, marked as a path on 1891 but not on modern maps, then there are near continuous boundaries through to Tan y Marian which are not accessible by public path. At SH62407089 the line crosses a by road which climbs very steeply up the hill to S through a small valley, and represents a drove road from Anglesey via Aber Ogwen through the Carneddau. At SH61987057 there is a clear terraceway by the field boundary, there used to be a house here according to 1838, which may be why the adjoining wood is still called Tan y Marian Bach on OS. A footpath comes up from Talybont onto line at SH61687022 and becomes a terraceway in edge of wood opening out to a walled track turning away from the alignment and going to a T junction at SH61626987. The track it meets seems to be a variant of the current minor road from Talybont to Llanllechid. There is then a FP over a footbridge over a stream at SH61606972 which is clearly not now on the line of any “road” but may be a memory trace? Coming up from the bridge to the present Llanllechid road there is a minor road opposite going to Tyddyn Isaf farm. What is now the accommodation road for Tan y Marian remains on the alignment, but of itself looks modern, and there is no obvious continuation on the opposite side of the Llanllechid road. This area needs more interpretation, the stream though small is quite a sharp drop and further W becomes marshy, which may help explain the lack of clarity.

At SH60976968 it looks as though a hollow way coming down to a marshy area has been “adopted” by a stream, then there is a track with an old quarry on S side going along field boundary to road at SH60816953, which is very nearly on the same alignment as that approaching Tan y Marian. Beyond the road is the well-known C15th manor house of Cochwillan (spelt as Cochwinllan on 1891), and on the far side of this at SH60626938 a blocked path goes into Coed Cochwillan, a wet hollow at first becoming a terraceway and taking a zig zag down through the wood to Afon Ogwen. The riverbanks are quite easily negotiable hereabouts, SH604693 but very steep to N & S, so it would be a good place to cross. Evidence there used to be a road and bridge crossing this stretch of river is provided by GAS XQS/1660/164, which is a presentment for repair of an unnamed bridge over the river Ogwen, flowing between Corrion (Cororion) and Bodvayo (Bodfaeo), situated on the road between Bangor and Conwey. Bodfaeo being the name of a mediaeval township encompassing Cochwillan.

On the far side of the river the access track to Lon Isaf could represent the line, then at SH60206911 we cross Ws road/Hen Durnpike from Capel Curig to Penrhyn Park, and a very narrow lane continues nearly on alignment with a small zigzag to get up a steep hillock at SH60076897. Just short of Cororion the lane turns S, and an accommodation road continues to the farm. Beyond this a terraceway can be seen at SH59526853, then a double boundary fence continues through Cororion Rough for nearly half a mile, both on 1891 and at present. After a slight gap and nearly on the same alignment at SH58956820 there is an access road to the house Pont Felin Hen (ie old mill bridge – there is an unnamed stream here). If you access the disused railway line, now a cycle route, between these two points, a wicket gate which would lead to the access road can be seen in the wall to the W, SH59046826, opposite on the E wall a corresponding access point has clearly been blocked with stone work, this strongly suggests that a right of way originally came through here.

From this point the B4366 is now the road leading directly to Caernarfon, but although the route is a direct one keeping to high ground, the actual road is a modern one based on a turnpike, last upgraded in the 1970s. The significance of the route in general terms is that it ignores Bangor, which has of course been a church site and subsequently a town since the C6th century (see variation of route above), and as M notes it is geographically well placed generally on higher ground between river valleys with commanding views. First however the Afon Cegin has to be crossed, the modern road does so near Groeslon roundabout SH56236685 but by this time the valley is so shallow that the crossing is imperceptible.

There does seem to be an alternative route on the northern side of the river represented by a disused track and right of way running from Tyn y Friddth SH58176841 (where it is now diverted from the front of the house N to the rear) past Tyn y Friddth cottage to Pen y Cefn at SH57436795 which has evidence of construction, side ditches, and some culverts. For this to be indicative of the Roman route we would have to postulate a crossing of the Afon Cegin at around SH585682, by way of a dogleg from the previous alignment. Topographically it would be a perfectly viable crossing, and once Tyn y Friddth is reached the advantage of higher ground as noted by M is gained. It also comes closer to the site of the now lost milestone found on Ty Coch farm (HH pp35-36). At Pen y Cefn the track ends on a minor road, there is a field boundary going straight on for about 600 metres which aligns well with the route now established for certain by aerial photography and excavation from Pentir switching station on to Caernarfon: most recently by excavation in advance of trenching a new pipeline at SH55596712, which I saw on 30th April 2021 when it was being excavated by Brython Archaeology.


1818 Ordnance survey drawings 2in/mile in British Library cat nos 301,305,306.

1838 Ordnance survey 1in/mile 1st ed

1891 Ordnance Survey 6in/mile & 25in/mile 1st edition 1891

GAS refers to Gwynedd Archive Service, Caernarfon

GATrefers to Gwynedd Archaeological trust records.

HH Edmund Hyde Hall, A Description of Caernarvonshire (1809-1811)

Hemp W. J. Y Cymrrodor vol33

M I. D. Margary, Roman Roads in Britain 1st edition

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

John Byde. Revision date 7th April 2022