The landscape impact of the construction of Roman roads is clear in parts of West and South Yorkshire, as these cut across the line of many existing field systems at Hook Moor, South Elmsall, Robin Hood’s Well, Burghwallis, Adwick-le-Street, Bramham and Rossington (Riley 1980: 94-5; Roberts, Deegan and Berg 2010: 58, 71-2, fig. 90). Some Roman roads even truncated earlier enclosures as near Hesley Hall, Rossington (Roberts, Deegan and Berg 2010: fig. 75). These roads were projections of Roman imperial power, disrupting familiar routes and routines, ignoring local tenure and tradition, and creating new politics of movement (Chadwick 2016b; Forcey 1997; Witcher 1997). Such roads allowed Roman soldiers and tax officials to penetrate deeper into the countryside, whilst the forced labour of local inhabitants was sometimes used to maintain them (Given 2004: 54). People living alongside roads could have their livestock, wagons and food requisitioned by Roman soldiers or officials, or face demands for hospitality and accommodation. This may well have caused suspicion and resentment. Yet such roads would also have linked families and communities more effectively, allowed livestock and goods to be taken to markets more quickly and efficiently; and would have speeded up the transmission of news. There would thus have been benefits as well as drawbacks. Investigating how Roman roads articulated with pre-existing trackways, fields and settlements can be a focus for future archaeological research; as can examining in detail where routeways and field systems re-orientated themselves to the new regime of movement.
Although the main outlines of most of the principal Roman road routes in South Yorkshire are known (e.g. Roberts, Deegan and Berg 2010: 68, fig. 88), there is much that remains uncertain. Analyses of cropmarks, lidar and geophysical survey data, and the work of the Roman Roads Research Association and Time Travellers organisations has provided valuable information to compliment or correct details of the major surveys undertaken by Margary (1973). The dates and chronologies of each road are still largely unknown, and a biographical approach would be productive. Though some Roman roads might have been initially used by the military for the invasion and occupation of the north, it might have been decades before they had full metalled agger surfaces and any flanking ditches constructed. Conversely, some minor routes might have remained largely as tracks with little or no metalling.
Roman roads are especially beloved of more traditional approaches to Roman archaeology, and arguments about their routes seem to arouse strong passions. It might not be too heretical to propose however that Roman roads in themselves are of limited intrinsic interest. Instead, their archaeological value and the reasons for identifying, investigating or preserving them comes from the information they provide on the development and sequence of landscape features over time, the chronology and tempo of military and official administrative activity, the physical, economic and social connections between communities, and how settlements may have flourished and failed. Some of the major routes in South Yorkshire are outlined below.
Lincoln to Doncaster
The road from Lincoln (Lindum) to Doncaster (Danum) (Margary 1973: RR28a) crossed the River Trent at Littleborough-on-Trent (Segelocum)in Nottinghamshire and then extended north-westwards towards Bawtry where it is partly visible on lidar (http://roadsofromanbritain.org/gazetteer/yorkshire/rr28a), initially spanning the low-lying and marshy land on the River Idle floodplain on a corduroy causeway built of logs, brushwood and turf to the north of the later Roman fortlet (Kennedy 1984; Van de Noort et al. 1997). This was later replaced by a more solid metalled agger surface on a slightly different alignment, reinforced with rows of substantial timber piles to carry it relatively high across the peaty floodplain. One of these piles was dated to the 3rd century AD. After several probable repair episodes, the stone agger ceased to be maintained and eventually collapsed, spreading laterally across the surrounding peat surface. Samples taken from twigs in peat above this spread in order to try and date the abandonment returned an Iron Age date of 180 BC–AD 50 (1σ) or 360 BC–AD 120 (2σ), suggesting that much older material had been redeposited by River Idle flooding (Bayliss et al. 1995: 88-9).
Entering under modern Bawtry somewhere under Queen Street Crescent, much of the alignment of RR28a is today marked by the line of the Great North Road or A638 leading past Gally Hills, King’s Wood and Hurst Plantation (Margary 1973: 410-11), with a possible deviation of 60m close to the Northern Racing College, and passing c. 500m east of the vexillation fortress at Rossington Bridge. It crossed the River Torne to the east of Rossington Bridge and Rossington Bridge Farm. Around Rossington and Cantley there might have been several Roman road junctions, perhaps indicating several different phases of construction or changes in alignments (http://roadsofromanbritain.org/gazetteer/yorkshire/rr28a).
The line of the road to Doncaster was traced north-west by Margary past Bessacarr Grange, along Cantley/Sand Lane and the edge of Doncaster racecourse, and entering Doncaster under the modern A638 (Margary 1973: 411-12). This route proposed by Margary has been criticised for being rather indirect and unlikely (http://roadsofromanbritain.org/gazetteer/yorkshire/rr28a; White 1963). Instead, using evidence from 1st Edition Ordnance Survey maps and lidar, it is suggested that there was a crossroads just north of Rossington and/or at Cantley, with RR28a then a straighter route across Doncaster Carr and under the modern suburb of Belle Vue. This line may be preserved by Roman Road and South Parade in Doncaster, before linking up with Hallgate and High Street, then Frenchgate and a presumed bridge across the River Don (Buckland and Magilton 1986: 30, 49, 55, 208; Chadwick and Burgess 2008: 122).
Another road, the so-called Cantley Spur (Margary 1973: RR281), headed NNW along the line of Warning Tongue Lane, past Cantley Hall, Sandall Beat Wood and Armthorpe, towards the fort long suspected and now confirmed at Long Sandall on the River Don. It is still visible in places as an earthwork (Roberts, Deegan and Berg 2010: 68). This might have even been part of a larger, abandoned scheme to link RR28a from Bawtry northwards to the fort at Roall and across the Humberhead Levels (http://roadsofromanbritain.org/gazetteer/yorkshire/rr28a). Lidar imagery also suggests that there might have been a short spur along the northern edge of the vexillation fortress and linking it to RR28a.
Analysis of aerial photographs has identified a previously unknown probable Roman road nearly four kilometres in length that extended almost due south from the south-eastern corner of Rossington bridge fortress, but which was unknown to Margary or Riley. It too apparently cut across pre-existing field systems and trackways, and the enclosure near Hesley Hall, Rossington (Deegan 2007: 25; Roberts, Deegan and Berg 2010: 68, figs 75, 88, 90). If it continued southward for 5.5km on this same alignment it would have met the crossing of the River Ryton just north of Blyth. This has been numbered as RR282 and may represent an alternate route for RR28a from a different phase, a northwards spur from a road from Lincoln to Templeborough (RR289), or even a route that eventually joined the RR5f Fosse Way to Willoughby (Vernemetum) and Leicester (Ratae Corieltauvorum). This might have extended northwards from the Rossington bridge fortress to join RR28a near Cantley, where an ‘Old road’ marked on the 1854 1st Edition OS map and lidar provide hints of a route (http://roadsofromanbritain.org/gazetteer/yorkshire/rr28a). These various routes were probably constructed in different phases – it is notable that the vexillation fortress at Rossington Bridge was not aligned to the known route of RR28a but appears to be more in line with (and at a possible bend in) the recently discovered RR282. More detailed lidar analyses and targeted excavation coupled with scientific dating techniques are required to resolve some of these questions.
Doncaster to Burghwallis, Castleford and Tadcaster
This proposed Roman route north-westwards from Doncaster – the Roman Ridge (Margary 1973: 415-6, RR28b) – also has problematic aspects. It survived as a prominent earthwork feature and was recorded by Daniel Defoe in the 1720s and Codrington in the early 20th century but has since been much denuded where it is not directly underneath the line of the modern A1. It appears to have been laid out as a series of straight but kinked sections skirting the low-lying and boggier ground in many places such as that east of Bentley and Adwick-le-Street.
After leaving Doncaster, the road’s exact crossing place over the River Don is unknown but may have lain under the present A638 and it then extended through Sunnyfields and Scawsby (Buckland and Magilton 1986: 208; Margary 1973: 415), continuing as a tarmacked bridleway and byway. After Green Lane it continues today as an elevated minor lane and causeway past Highfields and Adwick-le-Street, changing direction slightly several times. Several sections excavated across the presumed alignment at this point in advance of cycle route construction found remnants of a road surface at the southern end, but by Hanging Wood the supposed agger seemed to be a low bank some 25 metres east of the Roman Ridge (Tinsley and Pollington 2010). The ‘bank’ might therefore have been a medieval boundary or trackway parallel to the original Roman road. At Red House Farm, Adwick-le-Street, an unpublished excavation by Dorothy Greene in 1957 during dual carriageway construction recorded a well-preserved agger, and pottery from a roadside ditch or burial cist comprised early to mid-2nd century wares (Buckland and Magilton 1986: 193, 195, 208, fig. 42.351-2).
At Redhouse Farm, RR28b was only c. 60m away from an Iron Age and Romano-British enclosure (Area 7 Enclosure E1). Two sections excavated across it in the early 2000s found no surviving original road surface but the makeup layers of the agger were 8–10 metres wide and up to a metre thick. These consisted of layers of crushed stone and clay, with flanking sandy deposits. Beneath these were found different sets of furrows. Some were from pre-road cultivation, but others were deep, stone-filled features that were probably part of Roman road construction practices as outlined by the Roman writer Statius (Meadows and Chapman 2004: 11-15, figs 8-9). The Roman road cut across earlier field ditches, and one of these ditches produced an un-urned and undated cremation burial east of the Roman road (Upson-Smith 2002: 14). This important excavation remains unpublished.
At Robin Hood’s Well near Burghwallis, the modern A1 curves slightly to the east, but it is likely that the line of the Roman road was originally on a straighter alignment (http://roadsofromanbritain.org/gazetteer/yorkshire/rr28b). None of the three possible Roman forts immediately to the east were orientated to the road, which may indicate they pre-date it. How and exactly where the road crossed the low-lying valley and The Skell beck in unknown, though this crossing may lie beneath the modern upgraded A1. After climbing the slope on the other side of the valley, at Barnsdale Bar the Roman road changed orientation to the north-west, where today it mostly lies beneath the line of the modern A639 Roman Ridge – though sometimes it is adjacent to and parallel to the modern route. After crossing a sloping shoulder just north-east of Thorpe Audlin, the RR28a descended into the valley of the River Went.
In 1982 the Pontefract and District Archaeological Society excavated sections across the road north of Thorpe Audlin but on the south side of the River Went within South Yorkshire, immediately east of the modern A639. Their trenches revealed furrows or ard marks beneath the agger similar to those at Adwick-le-Street and likewise probably to prepare the ground surface prior to construction. The road surface was formed from sandstone slabs with the agger comprised of layers of sandstone and Magnesian Limestone rubble. There was only a roadside ditch on the western side, as on the eastern side of the road the ground sloped away steeply. A human shoe or sandal print was found beneath the agger layers, and a clay surface between the roadside ditch and the western camber of the road had preserved hoofprints of horses or cattle (Houlder and Hedges 1982: 3-4). The Roman road appeared to have been originally 5.5 metres wide, but this was later extended to c. 9m.
The agger of road RR28a by the River Went is still apparent in Google Earth aerial imagery and is approximately 135 metres north-east of the probable fort – the Roman road must have crossed the River Went only 30m east of the modern A639 bridge. Once again, the Roman road was on a different alignment to the fort which may therefore have pre- or post-dated it. The late date of pottery recovered from a section across the possible fort ditches may suggest the latter (Houlder and Hedges 1987: 2). A stone wall, a possible stone working surface or workshop floor and an iron roasting furnace that were also excavated (Houlder and Hedges 1982: 3-4) may have belonged to a vicus or roadside settlement. This important evidence remains unpublished. It has also been suggested that a possible Roman road RR18f from Templeborough might have joined the Roman Ridge near the fort (http://roadsofromanbritain.org/gazetteer/yorkshire/rr28b).
After the River Went crossing the Roman Ridge changed alignment slightly and extended north-westwards until East Hardwick in West Yorkshire, with the road either under or immediately alongside the A639. It then diverged underneath Sandy Gate Lane and alongside Hundhill Farm and Potwells Farm. A Roman milestone was found at Hundhill Farm, and geophysical survey and excavation at the findspot by the Pontefract and District Archaeological Society in 2002 found the largely plough-removed agger and a roadside ditch (Ferguson 2003; Houlder 2003). The inscription on the milestone has been transcribed as IMPC MANNI OFL[..] RIANO PF AVG INV MP E XXVI, which can be interpreted as ‘Imp[eratori] C[aesari] M[arco] Anni|o Fl[orliano] p(io) f(elici) Aug(usto) Inv[ictus], or “For the Emperor Caesar Marcus Annius Florianus [Pius Felix Augustus], Augustus Undefeated, Miles to York 26” (Tomlin and Hassall 2003: 371). Florianus only ruled for a few months in AD 276.
Further to the north in West Yorkshire between Micklefield and Aberford, near Hook Moor, sections along and across a steeply embanked section of the Roman Ridge road were excavated by AS WYAS as part of the M1–A1 Link Road project. An enclosure and boundary ditches pre-dating the road and likely to be of later Iron Age date were identified, which were later incorporated into a broadly co-axial field system featuring a prominent NWW–SEE aligned double-ditched trackway up to 10m wide (O’Neill 2001: 108-111, figs 85-6). A series of north-south ditched boundaries were arranged on either side of this trackway. Cropmarks and geophysical survey indicated that several subrectangular enclosures or annexes were appended to this trackway (Deegan 2001c: 33, fig. 17), although only one was excavated. This ditch contained animal bone, flint, industrial debris, metal objects including a flesh hook and 2nd century AD Romano-British pottery. A 14C date of AD 34–242 from material in a ditch fill is likely to be in the earlier part of the range. Due to truncation by later quarry pits only a few internal features were identified within the enclosure, but animal bone, slag, cinder fragments, clay hearth lining and uncharred barley grains were recovered from some of these (O’Neill 2001: 114). This enclosure would have been situated at the southern end of a flattish hilltop in an undulating landscape, and the exposed location may have assisted with the metalworking activities.
The quarry pits were associated with the construction of the Roman road, but the fact that some of them respected the enclosure indicates that this was an extant landscape feature when the Roman engineers arrived. Prior to the laying of the agger surfaces formed of alternating layers of subsoil and crushed limestone obtained from the quarry pits, two linear stone banks were constructed from weathered surface stones which may have formed markers for the road’s route (O’Neill 2001: 114-5, figs 88-9). A similar line of marking out stones was excavated at 8–10 High Street in Doncaster (Chadwick and Burgess 2008: 25-6). The overall agger width of the RR28a Roman Ridge could not be determined due to truncation by modern service trenches but was at least 7.2m and it survived to a height of 0.80m. No parallel roadside ditches were dug along this embanked section (O’Neill 2001: 115). Sandy layers that accumulated on either side of the agger may have been derived from windblown deposits, but like the clay deposits at Thorpe Audlin might also have been deliberately dumped to help the passage of unshod animals.
Brough-on-Noe to Templeborough
Perhaps the most perplexing putative route is the proposed Roman road between the fort at Brough-on-Noe (Navio) and Templeborough (Margary 1973: 361-2, RR 710b) which once again was suggested as taking a variety of alignments by antiquaries and early archaeologists. Even the initial route eastwards along the Hope Valley has proven problematic, as it was far from clear which side of the river this was on (Preston 1957; Welsh 1984; Wroe 1982). More recent analysis of lidar evidence now suggests that a line extending south of Bamford was most likely (http://roadsofromanbritain.org/gazetteer/yorkshire/rr710b).
One major flaw of the route proposed by Preston (1957) was that the road would have had to ascend through a steep and narrow cleft in Stanage Edge, which has always seemed unlikely – this is more probably a pack-horse route associated with post-medieval millstone quarrying. From there Preston suggested the road passed over Hallam Moors to cross Lodge Moor, and then along a route in front of Weston Park and the University of Sheffield to just below Park Hill where it then supposedly turned north-westwards towards Templeborough and met the Catcliffe to Oldcoates road RR189 headed east (Greene 1955, 1957a; Margary 1973). At Lodge Moor Preston’s investigations did reveal a metalled road surface nearly 9.5 metres wide with flanking ditches, but no real dating evidence (Preston 1957: 332-3). Welsh (1984) identified a possible stretch of Roman road extending from Scraperlow near Hathersage, passed near Carl Wark, ascended onto Burbage Moor and then descended towards Sheffield from Houndkirk Moor and past Ringinglow.
More recently, the broad outline of Welsh’s proposed route has been investigated by the Time Travellers archaeological society, who have analysed lidar imagery and conducted a series of geophysical surveys, earthwork surveys and small-scale excavations (http://roadsofromanbritain.org/gazetteer/yorkshire/rr710b). At Sheep Hill Farm, Ringinglow, excavations revealed a metalled road surface at least 12 metres wide with flanking ditches up to 5m wide and 1.3m deep (Inglis 2016: app. 4). This road seems to have extended as far as Limb Brook, where it might have been overlooked by the possible timber signal or watch tower on Bole Hill which was approximately 250-300m to the north-east (Waddington 2017).
From the Limb Brook the road has not been traced further but might have crossed Ringinglow Road just west of Firs Farm, and then along the length of footpath that links Cottage Lane to Common Lane. It may then have changed alignment to the bottom of the hill on Whiteley Wood Road, crossing the Porter Brook before climbing to Ranmoor (http://roadsofromanbritain.org/gazetteer/yorkshire/rr710b). it could then have continued in a more easterly direction to descend into Sheffield broadly along the modern line of Fulwood Road, meeting Preston’s preferred route through Sheffield and eventually Greene’s Catcliffe to Oldcoates road. Alternatively, the Roman road may have turned north-east in Darnall and followed the southern bank of the River Don to Templeborough. Leader identified a road extending eastwards from Templeborough to Park Hill in Sheffield (Leader 1878).
Margary suggested a line from Lodge Moor to the former Lodge Moor Hospital, and from there to Hallam head, Sandygate Road, Lydgate Lane, Hallam Lane and then down past the University of Sheffield and Sheffield Museum and Art Gallery (Margary 1973: 362). This part of RR710b is noted on South Yorkshire HER records as dubious, however, and more likely a medieval or post-medieval pack horse route.
Templeborough to Castleford or Thorpe Audlin
It has long been posited that there must have been a Roman road to enable direct movement from Templeborough further north to Castleford (Lagentium) by joining either RR28b or another road (e.g. Bishop 1999: 309), and antiquaries and older generations of archaeologists suggested (and contested) various proposed routes often based on supposition, place names and ambiguous earthwork or excavation evidence (e.g. Codrington 1903: 272; Greene 1950b: 168; Kitson Clark 1931: 186; Margary 1973: 413-4; Whiting 1931: 258-9). It was even provisionally numbered as 18f (x) by Margary and as RRX2 by the Ordnance Survey; and seen as essentially a continuation of RR18e Ryknild/Riknild/Icknield Street.
Aerial photographic analysis identified a c. 600m length of possible Roman road near South and North Elmsall in West Yorkshire (Deegan 2007: 25; Roberts, Deegan and Berg 2010: 70, fig. 90), orientated NNE–SSW and probably cutting obliquely across earlier field boundary ditches. Ahead of the A6201 Hemsworth to A1 Link Road project, a 2m wide section revealed two roadside ditches up to 4.7m wide and 1.22m deep with a central area of wheel rutting. No metalling was identified, and the ditch fills did not produce any dating evidence (Weston 2014: 24, 81, 177, figs 7, 19, plate 7). None of the visible cropmark relationships between the double ditches of the trackway or road and the field system ditches lay within the road corridor, so these potential stratigraphic relationships were not excavated or dated.
The projected line of the trackway or road extends to RR28b and the crossing over the River Went at Thorpe Audlin a few kilometres to the north, and it may be visible as a parch mark immediately west of housing in Thorpe Audlin on some Google Earth aerial imagery, though it is not apparent further south. The feature must therefore remain for the time being as only a possible Roman route.
Templeborough to Doncaster
This road was proposed as leading from east of the Roman fort at Templeborough and meeting RR18e Ryknild/Riknild/Icknield Street (Margary 1973: 412, RR710c) before extending to Rotherham broadly under the modern A6178. It might have then ascended to higher ground through Eastwood and to Dalton along the line of the modern A630 Doncaster Road, and then passed by or through Thrybergh, Hooton Roberts and Conisbrough, From Conisbrough the route might have broadly followed the southern side of the Don gorge before extending through Church Lane, Warmsworth and Littlemore Lane, Balby into the south-west side of Doncaster, just north of the modern A630. A slightly different route a little further south and mostly under the A640 might also be possible (Buckland and Magilton 1986: 209).
There is no archaeological evidence for this proposed road, however. A possible length of agger was investigated near Oswin Avenue School in Balby (Magilton 1977: 49; South Yorkshire HER 01859/01) but results were less than convincing. No roadworks or construction have revealed any traces of agger along the route. Near Hooton Roberts the A630 cuts across fields of probable medieval and post-medieval origin as parts of the Doncaster Road were constructed as the Tinsley and Doncaster Turnpike of 1764. The antiquity of the proposed route thus cannot be supported.
Other Roman road routes
Another proposed route is Margary’s suggested road from Templeborough to Littleborough-on-Trent (Segelocum) via Catcliffe, Oldcoates and possibly Bawtry (Margary 1973: 414-5, RR189), an extension of the road from Brough-on-Noe RR710b. This was in turn based on excavations ahead of opencast mining that claimed to have discovered metalled Roman road surfaces at Spa House east of Catcliffe up to 4.8 metres wide with an associated stone-lined ditch (Greene and Smedley 1955, 1957). This road lay underneath a late 17th century access road, but otherwise no dating evidence was found. Aerial photographic analyses identified two broadly east-west parallel cropmarks up to 530m long to the east of Firbeck Hall that might be the flanking ditches of a Roman road (Roberts, Deegan and Berg 2010: 70, fig. 88), but these features have not been traced further to the east or west so this route must remain largely conjectural until more research is undertaken.
Mention must be made of RR18e Ryknild/Riknild/Icknield Street (Margary 1973: 12-13). This Roman road extended from the Fosse Way at Bourton-on-the-Water in Gloucestershire and traversed most of the English midlands; and was still a noted routeway in the medieval period. From Little Chester (Derventio) it led to the Roman fort at Chesterfield, but north of there its course is uncertain, and Margary proposed two possible routes, RR18e and RR18ee, the former taking a more direct route and the latter following highest ground a little further to the east (ibid.).
The western route RR18e passed close to Woodhouse, Orgreave and Catcliffe where it supposedly intersected with RR189. This is all conjectural. From Bonewood Moor it then purportedly headed across Brinsworth Common just east of the site of Brinsworth Grange where Dorothy Greene claimed to have located a Roman road leading to the south-east gate of the fort at Templeborough, along with remains of a small Roman town (Greene 1950b, 1957a-c). Although Greene excavated several narrow trenches across stone road surfaces (Greene 1957a: 81-6; Greene and Wakelin 1950: 164-6), much of the line of the roads was established through probing or small test pits, and no artefacts were found in association with them. Some of the drawn sections of packed stones on rammed clay subsoil (Greene 1957a: 82-3) do not have the appearance of layered agger, although other sections are more convincing (Greene and Wakelin 1950: 166-7). This may be an old post-medieval coach road and other minor trackways of medieval or later date, and as Greene’s evidence for her proposed Roman town is unclear and has been discredited, the evidence for RR18e at this locale may also be suspect.
Greene’s evidence for a road approaching the south-west gate of the fort at Templeborough on the north-west side of Brinsworth Common is more plausible. Published plan and sections show excavated surfaces on the same alignment as the south-west gate and the road recorded by Leader and May, and the presence of roadside ditches and slight cambers in the surface is suggestive (Greene 1957b: 286-8, figs. 2-3). Again however, no dateable artefacts were recovered.
The second proposed route RR18ee probably extended from Chesterfield fort towards Eckington and is identifiable on lidar just north of the River Rother (http://roadsofromanbritain.org/gazetteer/yorkshire/rr18e.html), where an agger and a cutting are both visible. The road is visible on aerial photographs near Hagge Farm, where excavation revealed successive phases of metalled surface (Hart 1981: 92-94, plate 2), and later geophysical survey and excavation revealed more of the alignment (Derbyshire HER 13022). North of Eckington the course is unclear, but perhaps led through Beighton and Aughton. From Aughton it may have climbed onto high ground and joined with a possible Roman road traced by Inglis and the Time Travellers near Ulley, thereafter continuing through Aughton and Guilthwaite to Templeborough (Margary 1973: 413). It is odd that a Roman road did not follow the River Rother valley northwards, unless the low-lying ground was too marshy. Clearly, there are problems with the evidence for both routes, and further detailed research is needed.
Bridges, fords, valley crossings and wharves
Much research to date has focused on the Roman roads, but the places where they crossed low-lying valleys and rivers via bridges or fords is also worthy of future study. Although the lines of the corduroy causeway and later timber and stone causeway for RR28a from Lincoln on the River Idle floodplain just north of the later fortlet at Scaftworth was examined, the exact place where the road crossed the river is still unknown and remains of the bridge have not been recorded. Ground Penetrating Radar could be used to try and identify any timber pilings or stone footings below the existing banks. Just to the south, there was a detectorist find below or close to the early modern stone road bridge of a small Roman ‘purse’ hoard of eight 3rd century radiates, recorded in the HER and on Pastscape (Pastscape 321031). This could have been chance loss by a traveller but could also have been an offering prior to or following a river crossing (Bland et al. forthcoming), similar perhaps to the much larger deposits at Piercebridge (Walton 2016).
Despite the subsequent canalisation and dredging of river channels in the post-medieval and early modern periods, the crossing place across the River Don north of the fort at Templeborough, the full alignment of the crossing of the River Torne at Rossington Bridge, the line of the Roman road across the low-lying valley and The Skell at Robin Hood’s Well and the crossing of the River Went near Thorpe Audlin are all potential sites where geophysical survey and targeted excavation might be undertaken. Much may now lie underneath modern roads such as the A1 and A639, but it would be worth further investigation. Timber piles of unknown date were observed in the River Don near Wheatley (White 1963), and possible fords or crossing places such as one at Sprotbrough would also be worthy of future work.
The Roman forts at Long Sandal and Doncaster (and later the vicus) would have received many of their supplies by boats coming up the River Don, and this might also have been true of Templeborough. Bawtry was an important inland medieval port (Cumberpatch and Dunkley 1996), and so the fortlets and camps at Scaftworth were also probably part of a riverine supply route. No Roman wharves have ever been identified in South Yorkshire, however.
At Doncaster one location of such wharves could have been beneath the area under or around Doncaster College. Due to time pressures and budgetary constraints, the North Bridge/Low Fishergate excavations only proceeded to a depth of c. 3.5m below modern ground level, or roughly 5.8–6.6m above Ordnance Datum (McComish et al. 2010: 87). The ground surface was continually raised during the medieval and post-medieval periods by successive phases of building and demolition. Riverside timber revetments and a stone-surfaced draw-dock with stone walls were recorded, but these were medieval. Residual Roman material included coins, glass fragments and brick or tile. A 1988 borehole recorded ‘concrete, ash and brick fragments’ at depths of between 3.57 and 3.87m AOD (ibid.: 86) – the ‘concrete’ may be opus signinum. Roman-period deposits might also have been sealed by a 0.30m thick layer of alluvial silt representing late Roman or post-Roman flooding episodes (Buckland, Magilton and Hayfield 1989: 15). Residual Roman pottery was also found during the Doncaster Waterfront investigation near Doncaster College roughly 250m to the north-east. This work again found evidence of medieval waterfront structures, with residual Roman pottery suggesting features of this period might be present below unexcavated layers of alluvium (Brown 2004, Trench I).
The Romans in South Yorkshire may have attempted some large-scale engineering schemes as they possibly did in East Anglia and on the Gwent and Somerset Levels – though the nature and extent of the latter are still unclear. The canalised course of the River Don north of Thorne and the eastwards change in course of the River Idle to the River Trent at Stockwith created the Turnbrigge Dike and Bickers or Bycarrsdike canals (Buckland 1986: 40-42; Gaunt 1975; P. Jones 1995). These channels are recorded as having been cleaned out during the medieval period even if subsequently re-used by Dutch drainage engineers in the post-medieval period, and so may be earlier in date. Due to the later drainage and scouring work, however, attempting to date the first such digging would be extremely difficult.
Priorities and implementation