Ancient History & Civilisation

Chapter 2

Conquest and Construction

Existing winding trackways were sometimes utilized by the Romans, and were straightened somewhat and metalled. This is noticeable especially in Hertfordshire, the heart of the Catuvellaunian territory, where some parts of the roads have obviously been based on pre-existing Belgic tracks. The Viatores1

Roman Britain before Roman roads

So it was that when the Roman army landed on the coast of southern Britain in AD 43, they were immediately faced by a major logistical problem – how should they advance? Of course, this only seems like a problem from our point of view. The Romans themselves clearly had only limited choices, the most obvious and convenient of which was to use the existing native ‘trackway’ system. As we have seen, this need not have been as crude as has sometimes been assumed, but it is certain the army could not wait for even the simplest all-weather roads to be constructed, as this would have taken far too long. The presence of a large baggage train – both pack animals and wheeled vehicles – restricted the movement of the army, and travel without it could usually only be accomplished for a short period. Movement along the existing system of trackways and routes would have presented few problems during the campaigning season (traditionally March to September) and it would have been easier than travelling cross-country, although it would have meant the progress of the Romans could to some extent have been anticipated.

Surviving accounts of Roman armies on the march show how they sent detachments ahead with the specific duty of clearing the way for the main body of troops. They were evidently not meant to construct roads, merely improve existing ones. Indeed, the usual term seems to have been ‘to open up’ a road (aperire) and we find this being used of Tiberius in AD 10 and Germanicus in AD 14 when campaigning in Germany. In AD 9, the unfortunate Quinctilius Varus tried to march his army along inadequate tracks through dense woodland (some clearance was apparently required) and in foul weather, leading to the infamous clades Variana in the Teutoburg Forest. Not surprisingly, the use by the military of existing routes is attested elsewhere in the Roman Empire. In Arabia, theVia Nova Traiana followed an earlier Nabataean caravan route and presumably reflected part of the Roman advance into that province in AD 108.2

The formalization of a road network

So it seems that we must accept that the Romans did not, indeed could not, build roads at the same speed as they advanced. Properly constituted, all-weather roads, even if they were begun at the time of the invasion (and there is good reason to argue that there were better things for troops and engineers to do), would have followed far behind the marching army. Tacitus records how, when Germanicus was campaigning across the Rhine in Germany, the general despatched Lucius Apronius with a detachment to conduct road and bridge building behind his main force.3

Once they arrived in Britain, the Romans needed to establish an efficient infrastructure as quickly as possible. During the campaigning season (from spring to autumn) the existing tracks would have served, but once winter set in and the routes became wet and the rivers swollen, then lines of communications would definitely have been threatened. Constructing temporary ‘campaign’ roads along the native routes was one possible solution to this pressing problem and calculations published by John Peddie give some idea of the speed at which this minimalist exercise could have been accomplished. With the help of the Royal Engineers, he calculated it would have taken a force of 1,000 men fifteen weeks to construct one of his ‘assault roads’ from the Kent coast at Richborough to the Thames at (what was later to be) London. Such calculations (which would equate to approximately one yard per man per day) are, at best, ‘guesstimates’, but they nevertheless serve to provide some idea of what would be possible (and it has to be remembered that these are not proper surfaced Roman roads, but effectively cleared trackways). In reality, it is more likely that most Roman work of this kind was focused upon improving the existing road network, allowing the army to progress somewhat faster than Peddie’s figures suggest.4

An alternative system of roadways might have been provided by the use of log corduroy roads. This method was used very effectively by General Sherman to move his army through the Carolinas during the campaign of 1865, and he was reportedly able to achieve a rate of construction of twelve miles a day. We cannot completely rule out the possibility that the Roman army could have constructed similar roads de novo on occasion – there may even be a case for arguing that examples of such constructions survive beneath existing roads at Ambleside (Cumbria), Southwark (London), or Scaftworth (Notts) (see below, page 36) – but we can suspect that it was not their normal method of movement.5

The process of all-weather road construction during consolidation would need to have been followed rigorously as the Roman army advanced throughout Britain. In Germany, Germanicus realized the vulnerability of his lines of supply to adverse weather conditions and that was why he had Apronius work on road and bridge construction. It is extremely unlikely that forward bases could be maintained during the winter if they were not accessible by road, even if water transport was available. Later in the Roman period, outpost forts beyond Hadrian’s Wall and the Antonine Wall always had good roads linking them to their mural barriers.6

Campaign roads may be detectable in the archaeological record in some instances, although they will usually be concealed by later, more permanent roads. In Arabia, some broad paths were cleared through the basalt boulder fields of the Hauran but were never surfaced. They were marked by milestones, showing them to have been ‘constructed’ under the Emperor Septimius Severus, the cleared rocks piled on either side of the track. In Britain, parts of the Dere Street appear to diverge from an original prehistoric route which was only in use by the Romans between their conquest of Brigantia in AD 71 and the construction of its Roman successor, probably some time after AD 85. Likewise, the temporary camp at Rey Cross is bisected by the later trans-Pennine Roman road which clearly post-dates it, and this suggests that the force which constructed the camp – presumably in the early days of the Roman conquest of the north – were using an earlier trackway, the course of which was later followed (evidently not slavishly) by the new Roman road.7

Thus, it might seem logical that, once the campaigning season was over in the autumn, the army would withdraw to winter quarters and send out detachments to undertake road building on the grand scale. The Romans knew that idleness is the greatest enemy of the professional soldier, and road construction was one way of avoiding this problem. However, the Highland military roads of the eighteenth century were normally built during the summer season, usually between April and October. There is certainly good reason to suspect that, although Petilius Cerialis invaded Brigantia in AD 71, the final form of Dere Street was not constructed until at least AD 85, when Roman forces were starting to retreat from Scotland. This would allow fourteen years to construct 140 km of road between York and Corbridge.8

Practicalities of road building

Wherever possible, local materials would have been used. They could be obtained from most river valleys, and quarry pits can often be seen lining roads where the right geological conditions to supply gravel prevail (Plates 17 & 18). The timber used in the log corduroy between Scaftworth and Bawtry (S Yorks) included alder, a common tree in the carr landscape found in such a waterlogged river valley. Other corduroy-founded roads have been found on the south bank of the Thames at Southwark, approaching the Roman bridge to London and at Ambleside, in the Lake District, near the Roman fort.9

We have already examined Peddie’s calculations for the construction of ‘campaign’ roads, but it is much more difficult to guess how long a proper road would have taken to build. This is not only because we do not know the speed at which the Romans worked, but we have no idea of the proportion of their available manpower they might have been willing to devote to the task. We do have some brief details of how contemporary roads were constructed, thanks to the poet Statius (although even his account is sometimes dismissed as poetic fancy). Here he is describing the construction of the Via Domitiana, between Sinuessa and Puteoli:

The first task was to prepare a trench and open a route, excavating a deep hollow in the ground; the next to re-fill the open trench with other material, and prepare a base upon which the cambered surface of the road might be laid, in case the ground should subside or provide an unstable bed for deep-set blocks.Statius10

The picture gained from his short description can be fortified by analogy with the mural barriers of Hadrian’s Wall and the Antonine Wall. Both were built by gangs of soldiers, traverse the terrain in a similar way to roads, but – crucially – are accompanied with fairly detailed epigraphic records of their construction, even down to the amount built by each gang, allowing scholars to work out how they have been constructed. Using these as our guide, we can postulate how (and, importantly, how fast) a Roman road could be built. The mural barriers show clear evidence of having been constructed in stints assigned to particular units. Construction of a stretch of, say, Hadrian’s Wall might be allotted to a particular legion, and within that particular cohorts and, again, within that particular centuries. Although there was once believed to have been an overall movement of the construction process from east to west (albeit with possible variations), doubt has now been cast upon it being such a linear process. So, once a sector was constructed by the legions apportioned to it, they would have moved to their next allotted section and started the process all over again. A similar model may have been used in road construction, particularly for the major arterial highways, such as the Watling or Dere Streets. However, as was the case with the Walls, surveying and laying out need not necessarily have been undertaken in the same direction as actual construction. Nor, indeed, need the actual construction process have been sequential from one end to the other.11

We can also compare our conclusions with a later historical example of military road-building, fortunately better-documented than those of the Roman period. The subjugation of Highland Scotland in the eighteenth century saw a massive campaign of road construction under Marshall Wade and Major Caulfeild specifically aimed at facilitating the movement of troops. The military roads of Scotland were constructed in a similar way to the average Roman road: a broad trench with flanking drainage ditches dug, rubble foundations inserted, with finer gravel compacted on top. They were also constructed using more or less the same tools as were available to the Romans. Direct comparison is obviously difficult, not least because there would inevitably be special circumstances surrounding the construction of roads in highland terrain, but it can at least provide some hints as to what would have been possible in the Roman period. The expected rate of construction was 1½ yards (1.35 m) per man per day (at 16 ft – 4.8 m – most roads were just over 5 yards wide), and in at least one case 2 yards per man per day was achieved. This can usefully be compared with the Royal Engineers’ figures (of one yard per man per day) for clearing campaign tracks, but it is difficult to know where between these two sets of data one should place a Roman army detachment in terms of road building ability.12

At the sort of rates Wade and Caulfeild expected (and, as we have just seen, sometimes surpassed), we might hypothesize the 140 km of Dere Street between York and Corbridge to have taken approximately 285 man-years to construct. Taking this to extremes, if the entire British invasion force of approximately 40,000 men were put to road construction, they could produce slightly less than 21,000 km of road per annum, assuming they had nothing better to do. Absurd though this might at first seem, it at least provides a ceiling figure for the maximum achievable by the army in terms of road construction. It might be more reasonable to assume that a Roman commander might be willing to detail a proportion of his force – say a tenth – to consolidating his infrastructure. Thus, if his men were as good as Caulfeild’s, he might expect those 4,000 to lay 160 km of all-weather road in a month.

If work gangs were used to construct the road at several points simultaneously, this would obviously be quicker than one or two teams working from either or both ends towards each other. Comparison with the construction of linear barriers is obviously tempting here, particularly in the light of recent theories concerning the construction of Hadrian’s Wall (Hill noting evidence for several gangs at work simultaneously). Unfortunately, whilst construction records of linear barriers like the Antonine Wall preserve the distances built by particular units, they do not supply any details of the time taken and this has consequently become a popular academic guessing game. Wade worked with teams of around 500 men between April and October. Caulfeild increased these numbers substantially – in 1749 he had 1,350 men from five regiments involved in road construction (largely because he was building more roads than Wade), plus a further eighty on repair and maintenance. Tacitus refers to manipuli (pairs of centuries) being used to construct roads and bridges in the Balkans, prior to the army mutiny of AD 14, but it is not clear how much reliance should be placed on his choice of words (since it may just have been used in the sense of ‘detachments’). It might therefore be speculated that the 160 men of two centuries could construct 216 m per day at Wade’s rate of work, and thus about 650 days for the York/Corbridge stretch (or 478 days at Caulfeild’s rate). So, given a working season of seven months, it would take a force of that size three years to complete it.13


Before a Roman road could be built, as has been mentioned, its course would require surveying. Surveyors (mensores) had always been an integral part of the Roman army, primarily being involved in the layout of overnight and campaign camps, as well as more permanent bases, but also skilled in the construction of roads, bridges, and tunnels. It was their task to lay out the course of a road in such a way that it could easily be built by the army. The equipment they used is well-known and fairly basic: ten-foot measuring rods, flags (which served like ranging poles), and an instrument called a groma which allowed the siting of ninety- and forty-five-degree angles. They had no means we know of for measuring distance by line of sight (which requires precision optics or, increasingly nowadays, electronics). Nevertheless, all of the surveying operations necessary to construct a fully serviceable and long-lasting road were available to them.14

Although it might have been useful for the purposes of comparison, it is not known precisely how the eighteenth century Highland military roads were surveyed. Taylor has suggested that similar teams to those used by Watson and Roy during the Military Survey of Scotland would have been employed (an engineering officer, an NCO, and six men). We do know something of the survey that was undertaken in 1749 prior to the construction of the Military Road between Newcastle and Carlisle, since two copies of the plan survive, as do records of the operations of the surveyors. Two officers, Dugal Campbell and his deputy Hugh Debbeig, were assigned to the task and seem to have had enough time to note details of the Roman remains during their month in the country. Paradoxically, when the road came to be built, construction was contracted out to civil companies.15

Much is often made of the straightness of Roman roads (Plate 1). Many of the best-known do indeed have long straight sections, but most roads have frequent (if slight) course changes to allow for variations in the terrain they are traversing. What makes them so distinctive is that the sections of road – however short – are usually straight and not curved, unlike so many more recent British roads. This means that Roman roads make sudden adjustments of course of only a few degrees, rather than the sort of smooth gentle curves preferred for modern motorized transport. This is primarily due to the difference between a surveyed road (where its geometric origins are betrayed by angularity) and one formed by ‘line of desire’ (where lack of planning is revealed by a sinuous nature).16

Roads traversed long distances in this fashion, only varying from the pattern when they had to tackle a particularly demanding obstacle such as a river crossing. Often a sudden diversion would take the road down one side of a valley obliquely (sometimes down an engineered ramp), across the obstacle, then up the other side and resuming the old course, as happened at the crossing of the Haltwhistle Burn by the Stanegate in Northumberland.17

Interestingly, roads often maintained a general bearing, despite making deviations for various purposes. These were so-called ‘aligned roads’. It has been suggested that this implies that some form of long-distance directional surveying was used, perhaps utilizing the position of the sun or stars. Stane Street, as it leaves London, for instance, is said to be aligned exactly on the east gate of Chichester (Noviomagus), although such accuracy has long puzzled scholars. Some find it difficult to see how surveyors could have achieved this (particularly when the two endpoints were not intervisible), and some of the more ingenious methods proposed are less than convincing. However, if it is assumed that the road pre-dated the settlement, the position of which was thus determined by the road, then it all becomes much easier to understand. Crude cartesian coordinate systems were employed in the Roman period, as is shown by Ptolemy’s map of Britain, but it is not clear whether these were merely derived from a map or actually determined on the ground in some way and there is nothing to suggest their employment in the planning of roads.18

Of course, projecting a more-or-less straight line over long distances is relatively simple with even the most basic of surveying equipment – essentially all that is needed is a series of ranging poles along which the surveyor sights the alignment (a technique understood by Watkins) and it can be no accident that the major roads of southern Britain radiate from London; this would certainly seem to imply that the roads were surveyed from Roman London and might also be thought to give some indication of the fact that this did not happen until after that town became provincial capital during the first century AD. However, implicit in such an assumption is the notion that there were no pre-existing roads to use as models. Was a nodal point Roman or earlier?19

One interesting hypothesis has been advanced about the precise way in which Roman roads were surveyed in Britain. Davies advocated the use of maps by surveyors in their planning of routes to account for the accuracy of some roads in maintaining a general bearing (citing the example of Stane Street between London and Ewell pointing directly at Chichester). It is questionable whether sufficiently accurate maps were available to the Romans (certainly none have come down to us on a large scale that would fit the bill). More seriously, Davies does not allow for the possibility that there were already roads existing, which would conveniently provide the necessary (albeit approximate) alignments. To some extent, the hard work had been done for the Roman surveyors and all they had to do was cope with local requirements.20

With Davies’ scheme, the characteristic use of straight lengths is seen as a product of the way in which a road is surveyed and planned by transects, whereas earlier scholars have tended to opt for a ‘line of sight’ explanation, but both are, to some extent, flawed. In the case of Davies’, it is because of an over-technical reliance on planning and accuracy for which there is little supporting evidence; with ‘line of sight’, it has to be pointed out that for every course adjustment on a high point, there are a dozen that are not. If straight sections are a feature of planned roads, it is legitimate to ask why Wade’s and Caulfeild’s roads did not end up resembling Roman roads more closely (they often do, but not invariably). The missing element has to be the relationship to the pre-existing route, how it picked its way across the terrain (and most especially avoided obstacles which may, in the succeeding centuries, have vanished or changed) and, importantly, how it could be used to service the construction of its new, all-weather, straight-sectioned Roman successor.21

An alternative, and perhaps more practical, explanation has been advanced by Poulter. His examination of both roads and mural barriers (Hadrian’s Wall and the Antonine Wall) in Britain has suggested that the Romans often did survey (and, presumably, mark out) their alignments from hill tops, but that they tended to place their survey points on the far side of a crest. This more often than not allows their direction of survey (but not necessarily construction) to be identified.22

The course of a Roman road thus derived from several factors:

1.     overall heading would be provided by an existing route, almost certainly already in use by the army, and possibly even improved to some extent to prevent it deteriorating too much;

2.     the local course would then be a product of the line taken by the prehistoric route across the terrain, the Roman road approximating it (perhaps paralleling it, repeatedly crossing it, or a combination of these), but executed in straight lengths;

3.     particular topographical obstacles might cause slight adjustments in the line to avoid them, such as a dogleg to cross a watercourse or a zigzag to climb a hillside.

This still does not explain why straight lengths were used, and the answer here must be the same as it was for the mural barriers in northern Britain: it was easy for the army to accommodate. The Roman military surveyors could certainly cope with curves, both in terms of surveying and construction (each military base had at least four, one at each corner, and sometimes more in complex-shaped defences), but straight lines were fundamental to the army’s way of setting anything out. It would always be quicker to mark out an angular turn than it would be to describe an arc, and speed must have been a major consideration in laying out the new network. Moreover, an arc has a tactical advantage over a right-angle in castrametation; in road construction, such a benefit was absent and it would be an unnecessary flourish.23

Accuracy would always be a problem, since measurement could be haphazard in the Roman world. Without the benefit of precision engineering, small errors could occur in the transmission of the formal standards kept at Rome (in the temple of Juno Monetalis) by the time they reached the provinces. This situation was further complicated by the use of regional forms of measurement in some provinces – the league in Gaul and the pes Drusianus in various parts of the west including, it has been argued, Britain – and finally compounded by what appears to have been a slapdash attitude amongst the bureaucracy to the enforcement of ‘standards’. Thus, whilst a mile (mille passus) was always 1,000 paces, a pace (passus: technically two modern paces) was always five feet (pedes), the actual length of those feet was most definitely variable. The distance that is usually cited for a Roman mile is 1.48 km, but this can be found to range from 1.68 km to 1.89 km in Arabia (the last possibly due to a Roman surveying error!). In Britain, very few contiguous milestations survive, but in one or two places it is possible to reconstruct the original distances.24

On the Stanegate (Margary 85), the road that runs from Corbridge to Carlisle, three milestations survive near the fort at Vindolanda, two of these with in-situ (if truncated, in one case) standing milestones (Plate 4). Two more neighbouring milestations can be reconstructed and these seem to confirm a mile of about 1,550 m, implying a pes of 0.31 m – close to the pes Drusianus of 0.33 m. The next milestation to the east of Vindolanda – at Crindledykes Farm – includes a stone with the distance ‘XIIII’, presumed to refer to Corbridge which, even allowing for uncertainties over the exact course of the Stanegate to the east of the North Tyne, lies some 21 km to the east (and so close to a distance of 14 Roman miles, calculated on the basis of the Vindolanda mile of 1.55 km). A possible milestone was identified at Corbridge during the 1911 excavation there, the stone in question reported as being 2 ft 6 in (0.76 m) high and with a diameter of 1 ft (0.3 m), and two more milestones are known from the site, so it is possible that this is the point from which the XIIII was measured.25

On the Stainmore Pass (Margary 82), a group of three milestones at Spital in County Durham (so probably in situ at a milestation) were located 4.26 km from another stone at Old Spital to the west, equating to a distance of less than three Roman miles, which suggests the western stone has been moved, regardless of whether the pes Monetalis or the pes Drusianus was used. Sure enough, it was seen at a turnpike toll house in 1776 and may actually have come from slightly closer to Rey Cross camp. That camp derives its name from an early medieval cross erected within it. This was a marker stone, supposedly commemorating the site of the death of Eric Bloodaxe (the Norse king of Northumbria) or the boundary of the Diocese of Glasgow, originally located next to the southern side of the road on a small mound. Whilst there is no suggestion of a Roman origin for the stone, it is tempting to see it as set up close to a Roman milestation, from which the Old Spital milestone had been removed.26

More detail of British standards of mensuration can be gleaned from the Antonine Itinerary (see below, Chapter 3). This document gives distances between the various stations on each iter and thus allows us to compare the quoted with the known distance. If it is assumed that the distances in the document were measured by some means like the cart known as a hodometer mentioned by Vitruvius, which mechanically recorded distance travelled, then the error between the quoted and the known distances ought to be standard, at least within any given iter. In the past, scholars have assumed numerical discrepancies to be due to scribal errors in the manuscript tradition for the Itinerary (very easy with Roman numerals), but that does rather assume a standard mile (5,000 pedes= 1,000 passus = 1.48 km) was widespread in Britain. It seems at least as plausible to consider the possibility that not all Roman miles within Britain were the same length: mistakes might easily have occurred when recording the details on milestones, or even when laying out the road in the first place. Scribal errors are therefore just one of several possible causes of error in the recorded distances. Indeed, the spacing of the post-medieval turnpike milestones along Stainmore around Rey Cross suggest a mile of c.1650 yards being used in 1743, despite the fact that the statute mile had been declared at 1,760 yards in 1593, providing a curious coincidence with the possible Roman milestations and showing just how flexible a mile could be.27

Comparison of the Antonine Itinerary with the small surviving British portion of the Tabula Peutingeriana relating to the south coast provides some interesting detail (Table 1). This suggests, at the very least, that neither document is a copy of the other. Moreover, the tendency to underestimate distances appears to be present in both at some points, supporting the idea of some variability in the measurement of the mile.

Table 1. The Peutinger Table and the Antonine Itinerary compared

The Antonine Itinerary has even been used to argue for the utilization of a ‘Celtic mile’ of 1.665 km (as opposed to the usual 1.48 km), suggesting that this could explain the repeatedly shorter distances cited in the Antonine Itinerary than should have been the case if measured on the ground in standard Roman miles.28

Construction of all-weather roads

Varieties of construction types

Basic road construction in Roman Britain consisted of a number of components: foundations, surfacing materials, some form of drainage, and associated ancillary structures (like bridges) and roadside furniture such as milestones. The first stage in the actual construction of a road was to mark out the borders of the road by digging ditches and then stripping the topsoil. Statius, with a hint of poetic licence, describes this as the ‘excavation of a trench’. Opinions of his description of the construction of a road vary. Elsewhere, marking-out stones have been noted, during the excavation of the Roman Ridge (Margary 28) near Aberford (W Yorks) and in examining a vicus road at Inveresk. These consisted of stones set vertically before the foundation was laid and which, once the road surface was added, were no longer visible.29

An important element in the construction of a good Roman road was always its foundations. This normally consisted of coarse rubble or boulders, but under special circumstances other techniques were employed. Across low-lying boggy land, for instance, a log corduroy might be laid upon which the upper layers of sand and gravel would, in effect, float. An example of such a road has been excavated crossing the flood plain of the River Idle near Bawtry (see below, page 36). In the eighteenth century, Highland military roads were successfully ‘floated’ across boggy areas in a similar way, using metalling placed upon brush and log corduroys.30

The Roman roads of Hollywood films with neatly paved surfaces are a standard visual cliché, drawing on familiar images such as the Via Appia south of Rome or the streets of Pompeii. Such roads were not common in the provinces, not least because the resources were harder to come across, but they also required more work to produce. In Britain, there are a few suggested examples of paved roads, as at Blackstone Edge near Manchester or Doctor’s Gate, crossing from Brough-on-Noe to Melandra in the Peak District, but the identification of the surfaces (if not the courses) as Roman has been questioned. These had kerb stones and a central spine, with neatly-shaped setts like a nineteenth century cobbled street filling the space between, all enclosed within the customary ditches.31

By far the most common surface on a Roman road was one of compacted sand and gravel and such metalled roads are often found. The materials needed were easy to acquire (see below). In some areas, industrial waste was exploited to provide road metalling. In Sussex and the Forest of Dean, the debris from iron-making was used for this purpose and an example of a road surfaced in this way is still visible at Hoylte in Sussex.32

Regardless of the type of metalling employed, it was important that the road surface should have a camber on it to assist with run-off into the lateral ditches. Roads varied greatly in width and no ‘standard’ width seems to have been employed. The Stane Street was typically 6.45m wide, whilst Doctor’s Gate was 1.45m wide within its kerbstones, just wide enough for one waggon (perhaps another reason to doubt that it was Roman). The standard axle width was 1.4m, a fact that is evident from wheel ruts in the gateways of Roman forts and was a product of the most convenient width for a vehicle drawn by two animals yoked side-by-side.33

The various types of surface were primarily designed to take military traffic – the feet of the soldiers, the hooves of their pack animals, and the wheels of their baggage trains, as well as the infrastructure of the commissariat – so wear would have been light by comparison with modern macadamized metalling and the volume of wheeled transport which that has to cope with. The greatest enemies of a Roman road were probably the weather (with frost damage a major problem) and the encroachment of vegetation (especially on seldom-used routes). Milestones make reference to repairs conducted under various emperors and it is likely that the repair and maintenance of stretches of highway were the responsibility of the nearest town council (see below).34

Many roads that are found now have lost their original surfaces and only the foundations survive – this is probably the case on Wheeldale Moor (Plate 5) and in the vicus outside the fort at Brough-on-Noe in Derbyshire (Plate 6). Whilst the date of Wade’s Causeway on Wheeldale Moor has also been doubted, the road at Brough was undoubtedly Roman. The road leading out of the east gate of the fort at Inveresk was rebuilt (not just resurfaced) three times in the space of about twenty-five years, each time with a new foundation and new surfacing.35

Some roads have been found which do not conform to the model of fine surfacing material resting on coarse foundations. In these cases, metalling – often of fairly small pebbles – would be laid directly onto the existing subsoil (turf and topsoil having first been stripped away). Such a system would have the advantage of being very much faster to lay than a ‘proper’ road, less demanding in the amount of materials required, and adequate under most weather conditions. Examples have been noted at a number of places, notably Brough-on-Noe (Derbyshire) and Roecliffe (N Yorks) and it seems likely that this is a characteristic of early, conquest-period roads.36

Many Roman roads have their origins in major settlements like towns, but few actually pass through Roman military bases, it being far more normal for a site to be situated next to a road and a short branch road to lead to one or more of the gates. It is not hard to see why this should have been so: a large army marching along a road could have caused havoc by marching through the centre of a base, since it might take several hours to pass. Towns, on the other hand, would have been keen to attract visitors passing through. Corbridge is one of the notable exceptions to this, with the via principalis of the base forming the Stanegate once it left the western gate, which may well indicate that Corbridge was intended as the eastern terminus of the road (although it later extended further eastwards). Then, as Corbridge was developed from a fort into a town with a small military presence, things changed. Whilst Dere Street appears originally to have passed to the west of the castra, it was diverted across its old site and up the eastern side of the burgeoning town. At the same time, the Stanegate seems to have extended eastwards at least as far as the modern settlement.37

Once within settlements, roads very seldom had flanking open ditches but tended instead to use covered stone-lined drains and the Stanegate just to the west of Corbridge provides a perfect example of this (Figure 8).38

Figure 8: Sections through the Stanegate at Corbridge (from Forster and Knowles 1912).

The elements of a Roman road

The metalled surface

Crucial to the maintenance of a road was the prevention of standing water on its surface, and to avoid damage from running water. The first goal was achieved simply by the use of camber, shaping the profile of the road so that its centre was slightly higher than its edges. In most circumstances, water would then simply run off. Central stone spines – without a gutter – are also sometimes found, as at Wade’s Causeway on the North Yorkshire moors or on the Via Nova Traiana in Arabia.39

The margins

Nearly all roads were accompanied by drains along either side, serving to carry away the run-off from the camber. These might range from simple ditches to formal stone gutters, and perhaps even covered drains, as at Corbridge, the latter types especially in the vicinity of settlements. Roadside ditches were not universal, however, possibly because some well-drained soil types rendered them unnecessary. Ponding of standing water next to a road (which might eventually cover it) was undesirable, so culverts were sometimes set into the structure of the road and these have been noted on Wade’s Causeway. The road through the civil settlement outside the fort at Inveresk had no lateral drains, whilst that at Roecliffe did. However, the former was equipped with a culvert to carry water away and down the neighbouring hillside.40


Roman roads were typically marked at intervals of a Roman mile (however that was defined) by inscribed mileposts or milestones, set up at what are nowadays termed milestations. These are important as they mark the surveyed distance from a point of origin on a road. The provenance of a milestone is of course no guarantee of the proximity of a milestation, however. Since most have quite plainly wandered away from their original locations, those that have been found may also have done so. There are in fact at least four in Britain that can lay claim to being in their original positions: two near Vindolanda fort in Northumberland, one at Temple Sowerby in Cumbria, and one near Stinsford to the east of Dorchester in Dorset. We can to some extent rank the reliability of milestation identification: the least reliable is obviously a re-used solitary milestone, which may have been carried an unknown (and probably unknowable) distance from its original location; next comes a recumbent single milestone, usually excavated during construction or agricultural processes; finally, the most reliable has to be a group of milestones from one location. Even in the last case, the possibility cannot be excluded that milestones have been collected for re-use in the post-Roman period. One of the most plausible milestations in Britain has to be Crindledykes on the Stanegate (Margary 85), just east of Vindolanda. A group of five complete and two fragmentary milestones was found during field-drain digging here in 1885 (Plate 7). What makes it particularly noteworthy is that it lies only one mile east of the milestone opposite the fort at Vindolanda, and two miles from the in-situ milestone at Smith’s Shield.41

Milestones were usually (but not exclusively) cylindrical stone columns, often with square pedestals set into the ground. Fewer than 100 inscribed examples are known from Britain (Figure 9), but there must originally have been many more. Other provinces show how milestations could accumulate a succession of milestones (up to nine examples, at one milestation in Jordan), older ones simply being pushed over when a replacement was erected (Plate 8). At Gallows Hill, south of Carlisle, a stone with an inscription to the usurper Carausius was inverted and a dedication to Constantine placed on it, which suggests milestones were as much about display as they were about record and that it mattered whose name they bore. The seven Crindledykes milestone inscriptions have a date range from AD 222 to 337, although the Stanegate was originally probably built in the early second century AD. In a perfectly predictable world, this would equate to a mean of one milestone every seventeen or so years, so we might expect another eight milestones to await discovery at this site alone!42

Unlike the milestones of more recent times, Roman ones probably existed primarily to mark out the mensuration and did not always include an actual distance to a neighbouring destination (unless, of course, they were added in paint and have since disappeared). We do not know precisely how they were used: perhaps a tally was maintained by travellers wishing to record the distance covered or, perhaps, the army may have used them to judge their speed when marching. Some did give useful distances, however. One such example comes from Derbyshire, where the distance of 11 Roman miles to Navio (the fort at Brough-on-Noe) is noted whilst a stone from the Crindledykes milestation with a distance of fourteen Roman miles (‘MP XIIII’, presumably measured from Corbridge) has already been mentioned. The honorific inscriptions upon them record the names of the emperors under whom the construction or repair work was carried out, and most British examples belong to the third century AD (Figure 10). In Britain these are all carved, but painted inscriptions have survived elsewhere (Plate 9), which leads to the suspicion that some ‘uninscribed’ milestones, such as the one still standing near the fort at Vindolanda (Plate 4), may originally have borne a painted record. Moreover, there is an inference that can be drawn that carved inscriptions need not be primary, for there is now no way of telling whether they had been placed on what was previously a painted milestone that may, in its turn, have borne more than one painted inscription.43

Figure 9: The distribution of inscribed Roman milestones in Britain (based on location data in RIB I and III).

Figure 10: The dates of Romano-British milestones (Reece periods: 1: AD 1–41; 2: 41–54; 3: 54–68; 4: 68–96; 5: 96–117; 6: 117–38; 7: 138–61; 8: 161–80; 9: 180–92; 10: 193–222; 11: 222–38; 12: 238–60; 13: 260–75; 14: 275–96; 15: 296–317; 16: 317–30; 17: 330–48; 18: 348–64).

Thus the 100 or so inscribed milestones of Roman Britain probably present a highly biased picture of road construction and maintenance. Our dating for the milestones is, as we have seen, predominantly third century or later and may well be skewed. We can either assume from this that roads needed little maintenance from the time of their construction to the erection of the milestones or, as seems more likely, the practice of carving inscriptions on them only became popular in this period. This need not mean that they were not inscribed, since – as we have seen – they could have been painted.

Milestones were unlikely to travel far from their milestation after the Roman period, but manuports cannot be completely discounted. Since they were quite clearly an attractive resource for the acquisitive medieval builder, it is little wonder that most have now disappeared (Plate 10). Nevertheless, if every milestation was equipped with a milestone, then our 100 inscribed stones represent only 1.24 per cent of the original assemblage in Britain, based just on Margary’s suggested road network of approximately 11,938 km (8,067 miles). That, of course, can only represent a proportion of the original Roman network. If we then assume, on the evidence of Crindledykes Farm and milestations elsewhere in the empire, that more than one milestone may be expected at many British milestations, then that percentage will drop even further. At the very least, we may have to think in terms of upwards of 20,000 milestones throughout Roman Britain, in which case those 100 milestones form a mere 0.5 per cent.44


As was mentioned above, the junction of Dere Street and Stanegate at Corbridge has been excavated and was shown to have been of one build. This is not surprising, since the northerly course of Dere Street appears to have been diverted to pass through the settlement along the Stanegate, once the fort had been demolished in the second century AD. Roman roads tended to meet at an angle, if not at a nodal point such as a settlement. In such cases, there might sometimes be a short cut allowing a traveller to ‘jump’ from one road to another without having to go all the way to the apex of the junction. For example, traffic heading south on the Dere Street (Margary 8) could cut across to the Stainmore Pass by using Margary 820 to avoid going all the way south to Scotch Corner before heading north-west again (see below, page 43).45

Wayside shrines are still a common sight in Catholic countries and this is a practice that has almost certainly been continued from antiquity. Few examples have been identified in Britain (probably because they lie away from settlements and thus tend not to be excavated). Crossroads were thought to have supernatural associations and were the traditional burial place for suicides.46

It was a requirement of Roman law that all cemeteries should lie outwith the boundaries of a settlement. Cemeteries usually began by lining the roads outside settlements and then spread as the settlements grew and time passed. Many funerary inscriptions even appeal to passers-by and travellers, showing that proximity to roads was thought important for the dead and the need to remember them. Excavation on Roman burial grounds around London has shown how early ones were later encroached upon by the spreading town. The unusual barrow cemeteries outside the forts at High Rochester and Great Chesters in Northumberland carry interesting echoes of an earlier time and remind the visitor of the many prehistoric barrows that can be found close to Roman (and presumably prehistoric) roads.47

Bridges, fords, and landscape engineering

Modern authors have an inexplicable fondness for fords as an explanation for how Roman roads crossed rivers. In reality, Roman roads usually crossed rivers and streams by means of bridges; fords seem to have been much less common than they were in the medieval period (see below). Bridges were always preferred for the eighteenth century Highland military roads, not least as the passage of artillery was an important consideration, although in some cases the roads predated the final stone bridges (perhaps implying earlier, timber, temporary ones). Common strategic sense would demand that any river, or even stream, that was impassable when in spate would need to have been bridged, for as soon as a river became unfordable, it would have blocked the lines of supply and negated the whole rationale behind engineering a system of all-weather roads. This would almost certainly have been unacceptable for the army. The Roman military certainly knew how to ford a river – Vegetius gives a vivid description of how the cavalry would be strung out across the river to catch any who were swept away – but this was in campaign situations during the summer months when rivers would be at their lowest, and there were simply not enough cavalry in Britain to screen every river crossing in this way. Fords must have existed, but they will logically have been the exception, rather than the rule, where an all-weather road was required.48

So far as it is possible to tell, the most common Roman bridges were probably built of timber, so it is not surprising that few are known. Natural decay and the ravages of even the smallest watercourse can soon remove most traces of such structures, but there are exceptional circumstances where the remains of timber bridges have been found or where indications of their former presence have been noted. Even stone-founded bridges could have a timber superstructure, as appears to have been the case with Trajan’s Bridge across the Danube. The best known account of the construction of a timber bridge is that of Caesar describing the one he constructed over the Rhine. Trajan’s Column attempts to depict bridges of various sizes in its reliefs, including pontoon bridges, although its accuracy on such matters is questionable. Coins of the emperor Septimius Severus suggest that a pontoon bridge was used at least once in Britain, possibly either to cross the Forth or the Tay during his Scottish campaigns, and they may well have made sense on some tidal crossings.49

Where timber bridges crossed a stream flowing over bedrock, sockets for upright timbers can sometimes still be seen, as at Alfoldean on the Stane Street (where there were also traces of wood preserved). Squared timber uprights were also identified at Rossington Bridge. The discovery of the remains of a timber bridge over the River Nene at Aldwincle (Northants) has provided further useful insights into Roman bridge-building in Britain, but undoubtedly the most important find to date has been that of the Roman bridge over the Thames at London.50

At Aldwincle, a revetted, rectangular timber abutment 5.5 m wide has survived together with three phases of timber uprights, the whole structure apparently reinforced by angled buttress struts. There were indications that spans of about 3.5 m were used to cross the river. To the east, the road surface survived to approximately 6.5 m between ditches 10.5 m apart. The bridge evidently collapsed three times (at the end of the first century and again in the later second or early third century, then finally at some subsequent date). Each time it fell down, it deposited the actual road surface into the river, causing the level of the river bed to rise slightly at that point.51

Archaeologists have long sought the Roman bridge across the Thames at London, and strong hints were found to its likely location, but in 1981 the first major indications of its exact location were found during excavations on the Roman quayside at Pudding Lane. A substantial structure 7 m wide was revealed, which (it was thought) might be one of the piers of the bridge. The squared timbers interlocked to form a near-square box and at the time of its construction it was free-standing within the water (although later engulfed by the development of the quayside). The bridge probably employed both piers and piles to cross the full width of the Thames, for it has been pointed out that piering would not have been practical in the deeper parts. Constructed in the first centuryAD(after AD 78 according to dendrochronological dating), it was abandoned during the second century AD, probably in favour of a new bridge on a different site. It has been noted that there must have been both an earlier and a later bridge nearby and this seems wholly reasonable.52

A timber causeway, probably leading to a bridge, has been excavated at Scaftworth, where the main road from Littleborough to Doncaster (Margary 28a) crossed the River Idle, negotiating the marshy floodplain by means of a road surface laid upon a timber corduroy fastened in place with vertical piles. The bridge over the River Idle at Rossington Bridge has been examined in situ a few kilometres further north on the same road.53

It was thought at one time that the remains of the timber bridge at Newcastle (Pons Aelius) on the River Tyne had been found, but that idea has since been disproved and the precise location of the Tyne bridge remains uncertain, although the fact of its existence is not doubted.54

As in so much Roman construction, both timber and stone were used for bridges, although the former seems to have been more common on most Roman roads in Britain. There is a fine series of stone bridges associated with Hadrian’s Wall, two of which (Corbridge and Carlisle) carried important roads heading north (Margary 7 and 8), whilst others (Chesters and Willowford) carried both the Wall itself and the Military Way (Margary 86) across major rivers. There is some suggestion that these structures were – in part, at least – monumental in function, rather than just purely practical.55

References to stone bridges in Anglo-Saxon charters may in some cases refer to old Roman bridges. There is at least a possibility that some ‘-bridge’ place-names owe their existence to a surviving Roman bridge and Corbridge (which, it is often argued, did not receive a medieval bridge until the thirteenth century) may be one example of this practise. However, further along the Stanegate, near Chesters fort, the settlement of Chollerford lay next to the Roman bridge that carried Hadrian’s Wall across the North Tyne and that too had a medieval bridge so an Anglian bridge at Corbridge cannot be ruled out.56

The Romans did not just run their roads straight across the terrain but, where necessary, were quite willing to indulge in landscape modification. Where roads remain in use, such details are usually obscured or even obliterated, but the state of roads no longer in service can be more productive.

Roman roads often ran on an engineered embankment known as an agger and these are still to be seen in many places (Plate 11). They are not just restricted to low-lying or poorly-drained ground, as might be expected, and may have been due to the individual whim of the engineer responsible for any given stretch of road. It has even been suggested that insufficient excavation has been undertaken to be sure that some may not in fact represent an accumulation of successive surfaces, rather than one dump of material designed to raise the line of a road. Striking examples of an agger can be seen on Dere Street (now the A68), first as it heads south towards Corbridge (Plate 12), although part of it has been removed (possibly during the construction of the neighbouring turnpike road), and further north, at Ox Hills, where the modern road veers off the Roman course. A particularly impressive embankment survives at Roman Ridge, north of Castleford (West Yorkshire). The prominence of an agger may be one reason why later peoples showed a fondness for using Roman roads as boundaries.57

Roads that ran along the side of a hill would usually be terraced into it, which would involve cutting into the upslope side and depositing the spoil on the downslope side, thereby producing a platform for the road surface. At Haltwhistle Burn, the Stanegate is terraced into the hillside as it passes the fortlet, before continuing on the other side of the stream on an agger (Plate 13). Cuttings could also be used to cross a sudden break of slope, and may be found in approaching a river crossing. To illustrate the scale of some such operations, Trajan’s cutting at Pesco Montano (Italy) sliced through 38m of rock, requiring the removal of 13,600 m³ of stone.58

In mountainous areas, roads could be cut into the living rock, sometimes with ‘tramlines’ cut into the surface to help guide the wheels of vehicles. No examples are known in Britain as yet, but they are well-known in continental Europe – with examples at Bacharach in Germany and Langenbruck in Switzerland – and it is not impossible that some still remain to be found in Britain. Tight turns or hairpin bends were avoided by Roman roads except in such mountainous circumstances.59

So much for the ‘nuts and bolts’ of the road system. This has been covered by many writers before, but what is most significant about Roman roads tends to get overlooked: what can be deduced about their development and use and it is to this that we turn next.

If you find an error please notify us in the comments. Thank you!