Traditional accounts of Roman Cheshire, which have focussed on the nucleated settlements, have tended to  be descriptive in approach, while more recent studies of settlement dynamics in the area have been based on rural sites. This speculative essay attempts to reunite town and country. It  makes extensive use of current models of Roman demography and frontier society, as well as the results of recent fieldwork, to  explore three related topics: the size, composition and function of the canabae of the Chester legionary fortress; the development of the surrounding area as a supporting hinterland; and its social and economic trajectory. It is hoped to inform the research frameworks of future fieldwork, as a result of which the models will themselves be refined.

Introduction
Although what is known of the structural remains of the Chester canabae has been summarised several times, a detailed consideration of the human and economic dynamics of the settlement have been lacking: the make-up of its population, its roles vis a vis the fortress and the northern frontier and in the formation of local society. a detailed consideration of the human and economic dynamics of the settlement have been lacking: the make-up of its population, its roles vis a vis the fortress and the northern frontier, and in the formation of local society.  Indeed, overall reviews of the role of canabae in Britain as a whole are also lacking. However, the current model for military vici (ie settlements attached to auxiliary forts rather than legionary fortresses) is one of isolated communities of non-combatants dependent on the military units they supported, drawing on the neighbouring countryside for some supplies but failing to develop into market centres integrated into it (eg Jones, G D B 1984; Davies, J L 1991; Sommer 1984, 40, 50--2; Higham 1989; Evans 2005, 160). At Carlisle the level of occupation fluctuated rapidly, which McCarthy (2002, 80--1) has linked to changes in military dispositions. Indeed, Casey (1982, 126--7) suggested that vici were actually built by the army as an essential adjunct to forts; this hypothesis remains unsubstantiated, although in allocating land around forts the army does seem to have planned for their existence from the outset (eg Sommer 1984, 9, 11; 1999, 81--90).

The purpose of this essay is not to look in detail on the one hand at fortress--canabae  relationships or at the nature of the society of the canabae, nor on the other at the detailed patterns of settlement and rhythms of daily life in the hinterland, although these topics have perforce been addressed to some extent. A more complete assessment of the role of the Chester canabae vis a vis the fortress must await the publication of sites in the eastern extramural area, which has seen numerous excavations in recent years, alongside a reconsideration of older discoveries, while the quantity of information available for rural settlement in Cheshire, Merseyside and north-east Wales is still woefully small, and much yet remains to be learnt about the ‘small towns’. Rather, the present purpose is to focus on the relationship -- economic, demographic and administrative -- between fortress and hinterland and the possible role of the canabae  in mediating that relationship. A number of general models can be invoked to interpret the available evidence, and it is hoped that the attempt may stimulate some new avenues of enquiry. 

Models of Roman frontier economy and society 
The limits of expansion
Groenmann-van Waateringe and Whittaker have put forward related models to explain the limits to the expansion of the Roman empire as a whole, both of which draw on the research of Owen Lattimore on the Chinese frontiers (Lattimore 1962). In his ecological model, Whittaker (1989, 1994) argued that the limits of the empire became established in broadly defined, demographically and economically marginal zones beyond which a frontier defence force would be too heavily reliant on imported supplies -- and beyond which, by the same token, it would be profitless to expand. Marginality explains the importance of rivers, and indeed seas, to Roman frontiers, as supply routes rather than defences (Whittaker 1994, 85--97). For example, the Lower Rhine garrison was often dependent on supplies from elsewhere (cf Groenmann-van Waateringe 1989), while those on the Lower Danube and the eastern frontier both drew on the wealth of Asia Minor (Whittaker 1994, 99--100; see now Wilkes 2005, 144, 172, 213 for recent information on the former area). According to Appian (praef 5) the same was true of Britain -- ‘They have occupied the better and greater part of it but they do not care for the rest. For even the part they do occupy is not very profitable to them’ -- although this comment may reflect a stereotype rather than an understanding of the reality. However, the low carrying capacity of the highland zone of Britain has been disputed (most notably by Manning 1975; see now also Dark & Dark 1997, 113). Moreover, as Whittaker admitted, the creation of a frontier could itself be expected to stimulate local development. Groenmann-van Waateringe’s model posited that only societies with a developed hierarchy could be incorporated into the empire, as the Romans depended on the existence and cooperation of local elites politically to maintain their rule, and it was precisely in such hierarchical societies that the surpluses necessary to maintain an army were available (1980; cf Cunliffe 1988, 174; Millett 1990, 99--101; Whittaker 1994, 87). 

Turning now to Britain, Nevell (2001, 64--5; 2003, 11--12) has taken the view, based on an analysis of settlement sizes, that there was a very shallow social hierarchy in the north-west of England in the Late Pre-Roman Iron Age (LPRIA). ). Such hierarchy as existed in Cheshire may have been based in part on control of the distribution of salt. By contrast with south-eastern Britain, this did not require the existence of central places or the conspicuous consumption of durable material culture (Matthews 2000--1, especially 32--5; Nevell 2004--5, 14). White (2005) has expressed a similar view about the Cornovii, while Philpott (2000, 192) has pointed to the geographically peripheral situation of Cheshire within Cornovian territory and its apparently low population density in the LPRIA compared with the tribal heartland of Shropshire. These factors would have made Cheshire marginal in terms of the Romans' ability both to maintain garrisons and to manage local society and integrate it into that of the rest of empire. If these models are correct, then one would expect considerable development to have occurred around Chester to support the occupying forces. 

Taxes and trade
A particularly influential article in the study of the development of frontier societies has been that of Hopkins (1980), in which he argued that over c 50 % of empire-wide taxation was spent on the armies and that this would have constituted a massive redistribution of wealth from the inner ring of provinces to the periphery. Essentially this model develops an idea first put forward by Gren (1941, especially 147 ff) with reference to the Danubian provinces and taken up by MacMullen (1963, 89--95), A H M Jones (1974, 127) and A Birley (1981) and is a corollary of the economic marginality of the frontier zones referred to above. Some of this money would have found its way back to the core provinces to pay for necessary imports -- indeed, it might have been sought by their inhabitants to enable them to pay their tax bills -- but   one would also expect some to have penetrated into the societies of the frontier zones, giving rise to the local development identified by Whittaker and reshaping them into the more complex, 'Romanised' form summarised by G Woolf (1995).

However, we should note that Hopkins' article gave prominence to monetary taxes and considered their operation on an empire-wide scale. Other writers have given more emphasis to taxes in kind, especially in military areas, where they would have served to maintain the army (eg Garnsey & Saller 1987, 95--6, followed by Howgego 1992, 23; also Brunt 1990, 325--6, 531). Many examples of taxes in kind throughout the empire, not just in the frontier zones, were listed by Duncan-Jones (1990, 189--93). However, their importance was subsequently disputed by Hopkins (2002, 215--17). Likewise Mann (1985a), discussing Tacitus Agricola 19, argued that before the third century the Romans generally did not levy a tax in grain, except in areas where they had inherited one from former rulers: it was usually purchased. Nevertheless, he did note the exceptions of the Frisians, who paid a tax in ox hides, and the Batavians, who supplied recruits to the army in lieu of taxes, and wondered whether the same might have applied on the northern frontier of Britain. 

If taxes were raised in kind in frontier zones, government needs and the nature of local society may have coincided, as large quantities of agricultural produce were needed to feed the garrisons and the local economies were often initially unmonetised. Conversely, if taxes were demanded in cash, the government would have been the main purchaser of surplus produce sold to raise the necessary money: thus produce and money would have been virtually interchangeable (Duncan-Jones 1990, 193--4; Hopkins 2002, 216--17). 

The compulsory stoppages from soldiers’ pay in the late first--early second centuries  to pay for food and other supplies (see below) might be taken to imply that those supplies were purchased by the government. However, this need not be so: they could still have been acquired as taxes in kind which the government then turned into cash, by sale to its own soldiers and, perhaps through negotiatores, to emergent urban populations.

There is an important difference in the side-effects of taxation in kind and in money, in that the former would not have had the ‘multiplier effect’ of money taxes in promoting further economic activity and therewith social change (Hopkins 1980, 103; Whittaker 1989, 68--9, commenting on the failure of stable, urban-based landlord--peasant communities to develop in frontier areas). Moreover, even  if taxes were levied in cash, as long as transactions were restricted to sales of surplus produce to meet these obligations, the native economies would remain under-monetised; for further monetisation to occur, cash-based markets were needed (Casey 1992). These could have been generated, for example, by the supplementing of regular food levies for the army by additional purchases to make up shortfalls (Middleton 1979, 90--1) and by individual soldiers spending the portion of their pay which remained after stoppages on goods and services produced by the local rural population. In addition, unless they used requisitioned transport, traders engaged in long- and middle-distance supply of the army may well have wanted to buy merchandise from local civilians to sell on their return journeys (Duncan-Jones 1990, 42).

As we have seen, the army absorbed a large proportion of the taxes raised throughout the empire. These were arguably a heavy burden on the native population in the frontier zones: land there may well have ranked as public land (ager publicus populi Romani) with its occupiers having to pay rent to the state as well as tax (Mann 1985a, 21). By contrast, many of those related to the garrisons would have made no contribution. Up to the end of the second century not only serving soldiers but veterans, their wives, children and parents were exempt from direct taxation (tributum), indirect taxes (vectigalia) except that on inheritances, and from public services (munera). These exemptions seem to have been reduced in the Severan period, but that on munera remained (Birley 1981, 45--6, 49--50).

Models of settlement and society in the north of Roman Britain
The prevailing views of the impact of Roman military occupation on northern British society are based on studies carried out in Yorkshire and especially Cumbria. In the latter area N Higham and G D B Jones found apparently increased settlement in the Roman period (although very few sites were tested by excavation) in the form of dispersed farms. The average settlement density was one site per 3.6 km² overall, excluding mosslands and tidal areas. The size of lowland farms generally varied from c 60 ha plus rough grazing elsewhere to c 200 ha including rough grazing; the average size of site territories was probably c 100 ha, although it could be as low as 30 ha on the Solway eskers. Field systems were in the region of 6--15 ha. The remains clearly indicate mixed farming, with unsurprisingly a greater emphasis on pastoralism at the higher altitudes. The extent of cultivation in the Roman period had previously been underestimated, but later in the period there seems to have been a decline in arable production in favour of pastoralism (Higham & Jones 1985, 72--112). Forts seem to have been sited in productive areas from which they could easily be supplied, while the field systems attached to the vici could have produced fresh foodstuffs. As well as supplying the needs of the forts the vici also supplied the local rural settlements with traded goods, although only in very small quantities (Higham & Jones 1985, 109--10). 

However, in his 1984 paper Jones pointed out that while Roman occupation appears to have resulted in an increase in the density of settlement in Cumbria, it did not result in more specialised agricultural sites requiring markets, or in the more complex hierarchy of sites which emerged elsewhere in the north, for instance in the Vale of York: the vici remained urban islands dependent on the forts, with minimal links with the surrounding countryside (cf McCarthy 2002, 117; 2005, 58). Essentially the same conclusion was reached by J L Davies (1984, 1991) regarding the vici of the comparatively short-lived Welsh forts. Indeed, Higham (1989) characterised the Roman period in Cumbria as one of pauperisation, with the local population being deprived of the best land and government exactions further limiting the surplus they could sell in the few markets that existed near to the forts. The population of the vici was largely immigrant, and local people lacked the skills and resources to benefit from the exploitation of natural resources: indeed, a dual cultural system may have initially been officially encouraged. 

Nevertheless, part at least of the area ultimately achieved local self-government as the civitas Carvetiorum, centred on Carlisle (Rivet & Smith 1979, 301; Burnham & Wacher 1990, 54), so presumably a sufficiently large and wealthy class grew up to whom power could be entrusted. The existence of the civitas is now attested by three inscriptions, of which the earliest belongs to the reign of Severus Alexander, and it seems likely that it was established by Septimius Severus (McCarthy 2005, 49; on the earliest, most recently discovered, inscription, which confirms Carlisle as the civitascapital, see Edwards & Shotter 2005, 69). So far the existence of a civilian elite at Carlisle is attested from the late second century by tombstones, architectural fragments -- some possibly from public buildings -- land reclamation, the construction of an aqueduct and a couple of large houses with hypocausts (McCarthy 2002, 82--3, 88--90; 2005, 66). In the countryside, some circular houses were ultimately replaced by rectangular ones -- perhaps a sign of more extensive cultural change.

East of the Pennines increased dispersed rural settlement has also been found, eg west of Doncaster and in north Nottinghamshire. The emphasis on arable or pastoral farming varied: around Wetherby and Aldborough it was on arable; in the Yorkshire dales and in much of County Durham it was on pastoralism (Ramm 1980). In this region there is also evidence for some reorganisation of the landscape during the Roman period. For instance, fields seem to have been laid out per strigas, aligned on the Roman road, the Rudgate, south of the first-century fort at Newton Kyme (Ramm 1980, 34--5). 

In addition Branigan (1980) listed thirty-nine certain or possible villas in Yorkshire and County Durham. He identified the most likely builders/owners of these as wealthy veterans, traders and other ‘professionals’, and members of the native aristocracy, in that order. (Those of the native aristocracy are perhaps to be recognised by continuity of occupation from the LPRIA). The villas could therefore be expected to cluster around the military bases of York and Catterick, the fort at Malton, the civitas capitals of Aldborough and Brough on Humber, and the port of Bridlington, which would have provided security, markets and social centres. This does indeed correspond to the observed distribution, but the concentrations of villas in each group are small, as are the buildings themselves and, although many of the Parisian villas had mosaics, they were of poor quality compared with those in the south of the province. The Yorkshire villas were also generally of late date, with a floruit in the late third and fourth centuries. (For a recent  study of the distribution of villas in this region, see Burroughs 2001).

The construction and maintenance of these villas clearly demonstrates the concentration of surplus wealth. Especially where land changed hands, for instance coming into the possession of former soldiers, one might also look for a change in objectives to the Roman model of farming for profit (cf Percival 1976, 145--7). This is borne out by the location of villas near roads and markets and sometimes a change in field systems. In some cases, from the mid-second century, there seems to have been a concentration of the local dependent population. There was also an increased emphasis on arable farming. However, more recently, Roskams (1999) has pointed out the limitations of these changes. In particular, changes in field systems do not precede the third century.

Even if most of these villas were in the hands of outsiders, it is worth remarking that, by contrast with Cumbria, LPRIA Yorkshire had a better developed social hierarchy, Stanwick most probably being the capital of the Brigantian queen Cartimandua (Higham 1987; cf McCarthy 2005, 58--9. For the cultivation of larger areas of cereal using spelt wheat rather than emmer, with the surpluses supporting a more hierarchical LPRIA society, see Van der Veen 1991; Van der Veen & O’Connor 1998). In other words, the emergence of a villa-based rural economy in this region could be seen as a development of the pre-existing social order. It may also be significant that, as Evans (2005, 16--1) has noted, from the Antonine period onwards the vicus at Malton shows a high level of interaction with the neighbouring countryside in pottery and other goods. However, as Swan (2002, 71) has suggested, ultimately the key to the late period of prosperity may well have been an increased reliance by the Hadrian’s Wall garrisons on supplies from the area, exemplified by the prominence along the wall of east Yorkshire pottery. 

Finally, as is well known, in the early third century, York first became the seat of the governor of Britannia Inferior, c 212, and then by 237 its extramural settlement had been promoted to the rank of colonia, possibly having been previously designated a municipium, again c 212 (Wacher 1974, 156; Mason 1988b, 187--8; Ottaway 2004, 83).

Taking a broader perspective, two recent studies of the adoption of Roman material culture in Britain have re-emphasised the traditional view of a north-west--south-east divide, corresponding to Fox's (1959) highland and lowland zones (Sargent 2002; Henig 1999). Sargent drew attention to the well known lack of villas and temples in the north and west compared with the south of the country. However, he rejected the proposition that the disparity is to be explained by natural poverty, as the phenomenon is found (albeit to a lesser extent) even in environmentally relatively favoured areas near to the legionary fortresses at Caerleon, Chester and York. Nor, he argued, is it explained by the continued military presence stunting the growth of an economic and social hierarchy, as was argued by Millett (1990, 99--101), as the same lack is found near the short-lived fortress at Wroxeter. Rather, he adopted the view that the tribal elite in the highland zone chose to opt out of the 'Roman way of life' such as the construction of villas, and sought to display their wealth and status in other ways, eg through the ownership of herds of cattle and the support of clients. (For this view of LPRIA Cheshire, see Matthews 2000--1, 32; for Roman Cheshire, Nevell 2003, 16; for the Cornovii as a whole, White 2005). This is essentially an application to Britain of the model of Lower Rhine society put forward by Roymans (1995). 

Focussing on artistic production, Henig saw the highland, 'military' zone as being characterised by portable artefacts, the lowland 'civilian' zone by more durable production, for example mosaics, wall paintings and sculpture. Overall, the British limes stands in contrast to the Rhineland, where Cologne is notable for the quality of its mosaics and wall paintings and the quality of sculpture is comparable with that from Italy and Gallia Narbonensis. However, Henig also drew a distinction, at least in quality of execution, between Caerleon and York on the one hand, and Chester on the other. Caerleon was within striking distance of the Silurian civitas capital at Caerwent, the colonia at Gloucester and the flourishing towns of Cirencester and Bath, and its mosaics can possibly be paralleled by those produced by civilian studios. York shared a school of mosaicists with the Brigantian civitas capital at Aldborough; indeed it may have the home of the school. By contrast, he evaluated the sculpture produced at Chester as, with exceptions, mediocre to incompetent. It shared with much of north Britain -- where buildings are characterised by simple architectural decoration contrasting with the more exuberant work to be found in the south (cf Blagg 1977) -- the problem that there were not enough patrons to attract high quality craftsmen.

The Chester fortress, canabae and prata legionis: summary
Occupation of the fortress
The legionary fortress at Chester was founded at some time in the 70s of the first century AD. The conventional account is that it was Legion II Adiutrix which built and formed the first garrison of the fortress, but Mason has recently argued that this role was actually played by Legion XX VV, previously based at Wroxeter, II Adiutrix moving to Chester from Lincoln when Legion XX VV went north to take part in Agricola’s campaigns; the latter legion then returned to Chester to replace II Adiutrix when that legion was transferred to Dacia (2000, 167--76). Whichever legion initiated construction at Chester, the pattern of occupation over the next decade is likely to have complex. Coin evidence indicates that Wroxeter was not given up by the military until c 86/7, at the same time that Inchtuthil was abandoned (Hobley 1989). Between c 79 and 86/7 the headquarters of Legion II Adiutrix was probably indeed at Chester and that of  Legion XX may have remained at Wroxeter, but vexillations of both legions would have served in Scotland, and Inchtuthil may have been built by such a joint force. It was probably not until Legion II Adiutrix was withdrawn from Britain and Legion XX returned south that Chester was occupied in strength. We should also note here that Mason has argued that Chester may have been intended in the early Flavian period as the provincial governor’s residence, with the enigmatic ‘Elliptical Building’ intended as an Augusteum, but that with the advance into Scotland under Agricola it soon became clear that it lay too far to the rear for this purpose and a location at York or Carlisle far more relevant (2000, 91--5).

Legion XX probably had the largest share in the task of building Hadrian’s Wall, from AD 122 onwards, and part of it, at least, was still in the north between 128 and 138, when it was responsible for rebuilding the fort at Moresby on the Cumberland coast. Within a few years, detachments of the legion were again in the north, taking part in the construction of the Antonine Wall and then sharing in the garrisoning of southern Scotland. In the mid- to late 140s detachments probably also took part in the war in Mauretania. In the 160s it participated in the demolition of the Antonine Wall and the accompanying refurbishment of Hadrian’s Wall and many Pennine forts; a detachment of the legion was based at Corbridge, along with one from Legion II Augusta, and both remained there, and at Carlisle, in the third century. The large-scale absence of the legion from Chester in these decades is shown by clearly by the sudden interruption c 120 in the work of reconstructing the timber fortress in stone, eg on the Northgate Brewery barrack site (Ward & Strickland 1978, especially 27); this was not resumed until the 160s.

After taking part in Commodus’ suppression of the Armorican revolt, the legion is presumed to have contributed to the force with which Clodius Albinus faced Septimius Severus in 197 and then to have taken part in the latter’s Scottish campaigns (208--11). After the abandonment of Scotland by Caracalla most of the legion seems to have returned to Chester, where the fortress was subjected to an extensive rebuilding programme which extended over two decades (Mason 2001, 161--80).

Later in the third century we find evidence for detachments of the legion serving away from their base, not only in Britain, as at Maryport (238--44), but at Mainz (255); the latter detachment seems to have become part of a de facto field army serving in north Italy. Any detachments serving abroad are likely to have been cut off by the rebellion of Postumus in Britain (AD 260) and never to have returned. The latest datable inscription of the legion in Britain is a dedication from Hadrian’s Wall of the years 262--6. (For a brief account of the movements of the legion see Jarrett 1968; Swan 1999, especially 425--6; Swan & Philpott 2000, 63--4; Coello 1996, 18). Although the Chester fortress continued to be occupied, probably by the army, in the fourth century, it is possible that the garrison was no longer Legion XX VV. Petch (1987, 137), followed by Matthews (2003--4, 17) suggested that the legion was dismissed by Constantius following its support of the usurpers Carausius and Allectus, and recently Clay (2004) has argued that the defacement of many of the tombstones found in the North Wall of the fortress was not an accidental consequence of their reuse but was a deliberate act of collective  damnatio memoriae.

Thus documentary and epigraphic evidence suggests that the fortress at Chester was probably only held in strength for for less than three generations, c 90--120 and 210--260. Even during these periods, despite the lack of specific evidence, it is prudent to assume that there would have been times when detachments of the legion were absent. This is something we need to bear in mind when considering the impact of the army on the local area.

The prata legionis and canabae
Attached to legionary fortresses were the prata legionis, the ‘pastures of the legion’. Prata in general and at Chester in particular have been reviewed by Mason (1988b). This land could be held by the legion as a sovereign citizen body and, where the area of the prata can be calculated, could amount to over 500 km^2. It was probably used not only for pasturing the legion’s mounts and baggage animals, but also for keeping food animals ‘on the hoof’ and for the supply of raw materials. During the first century AD there grew up on the prata, immediately outside of fortresses, quasi-autonomous bodies of veterani et cives Romani consistentes ad legionem …(‘veterans and Roman citizens dwelling with legion …’) (see in general Mason 1987). These included the female companions, children and other relatives of the soldiers, and also traders providing essential supplies, prostitutes and entertainers, and servants and slaves (although the slaves of individual soldiers or units would presumably have been accommodated inside a fortress). Such a civilian ‘tail’ attached to the Roman army is already attested in Republican times (Maxfield 2002, 146--7). However, as the dispositions of the army became more static, their simple encampments became more substantial quasi-urban settlements, although their name -- canabae, ‘huts of the legion’, first attested in the time of Hadrian -- continued to reflect their origins. The inhabitants could probably only rent land (contra Sommer 1991, 475) and remained under the ultimate control of the praefectus castrorum, but nevertheless enjoyed a measure of self-government: cives Romani consistentes could appoint magistri or curatores, while canabae had decuriones and all the magistrates found in municipia and colonies (Mocsy 1974, 142).

In the early years it appears that the prata appropriated by legions were under-used and provoked resentment among the native population; later it is possible that they were occupied by farms let out by the army for rent. After the middle of the second century the term territorium legionis occurs, which possibly reflects a subdivision of the prata between the land occupied by civilians and that reserved purely for military use (cf Le Bohec 2000, 219--20). An inscription from Carnuntum suggests that the canabae may have been confined within a radius of one leuga (c 2 km) from the fortress (Wilkes 2005, 146, 198). Settlement evidence at Novae suggests that the same rule applied there (Conrad 2003).

In the absence of any direct evidence, eg boundary markers, Mason has suggested on general grounds that the prata at Chester possibly extended east to the mid-Cheshire ridge, south to Farndon, north to the Mersey and to the north-west included part or all of the Wirral. In the west, apart from a bridgehead at Handbridge, the boundary would have been formed by the Dee. This would give an area of up to c 480 km² (present author’s figures). The only known structures outside the canabae which would definitely have fallen within this area are the practice camps surrounding Chester to the north and east (Philpott 1998), although the farmstead at Irby (Philpott & Adams 1998) would also have done so if the whole of the Wirral was included, and there are hints of others (see below). 

In addition to the canabae there were often additional settlements sited c 2 km from fortresses. These were first recognised on the Rhine and Danube frontiers by Mocsy (see most recently 1974, 139--43, 217--25) and have been discussed in their British context by Mason (1988a, 137--54; 1988b, 176--8; Novae may now be added to his list: Conrad 2003). They may have been founded deliberately as purely civilian centres on land belonging to the local civitas rather than on the prata legionis. Such a settlement has been found near Chester, at Heronbridge, and is described further below. In the course of the second century the more successful of these settlements on the continent were granted municipal status and under Septimius Severus became unified with the neighbouring canabae, which thus acquired the same status. In the first half of the third century some were upgraded to the status of colony. It is thus possible that the Chester canabae and Heronbridge were united administratively as a municipium, most probably in the early third century, and that they acquired the former military prata as their territory, although whether the ‘Deva’ mentioned as the dedicators’ origoon inscriptions from Worms and Trier (Rivet & Smith 1979, 336) implies a municipal authority at Chester, or even refers to Chester at all, is uncertain (Mason 1985).

According to Ptolemy (II, 3, 11; see Rivet & Smith 1979, 336), writing towards the middle of the second century, Chester was a polis (city) of the civitas Cornoviorum, the latter being formally established in Flavian times (Wacher 1974, 358--9), and modern scholars conventionally assign the whole of Cheshire to that tribe. In fact, the fortress, canabae and the rest of the prata would all have been legally distinct from the civitas and it is most unlikely that Chester had any municipal status at this time. Further, given that some of the salt towns may have been imperial property (especially Nantwich; see below), one wonders how far the writ of a tribal council ran in any part of the modern county. It is possible that Ptolemy was writing loosely and merely wanted to show Chester’s existence as a nucleated settlement and its approximate location. 

In addition, while the Dee and Mersey estuaries and the mosslands of the middle Mersey could have served as boundaries of the Cornovii with the Deceangli and ?Setantii (a sept of the Brigantes) respectively (Rivet & Smith 1979, 331; 456--7), it is by no means obvious that the Dee above Chester also served as a high-status political boundary. (For example, Mason assumes that Heronbridge, on the west bank of river, lay within the civitas Cornoviorum). Moreover, the Wirral would also have been isolated from the rest of Cheshire by the wetlands of the Gowy estuary, and one wonders whether it ever lay within the territory of any known tribe. In fact, the conventional attribution of Cheshire to the Cornovii in any sense should be treated with caution. All these questions of legal status are important but are unlikely to be determined without the discovery of inscriptions or characteristic public buildings. 

The Chester canabae: function and population 
Function
A number of roles can be envisaged for canabae. These, of course, are not mutually exclusive.

1. Residential and communal areas for non-combatants, eg soldiers' households, developing into a recruitment centre for the legion, based on veteran’s families

2. Manufacturing settlements supplying the fortress with goods produced in the canabae

3. Service settlements living off traffic to the fortress

4. Service/trading settlements living off traffic en route to destinations further afield

5. Trading settlements providing fortresses with imported goods/supplies and others produced in the immediate area

6. Trading settlements supplying surrounding settlements with goods produced in the canabae, the local area, or imported 

7. Social and political centres for the local elite

Although detailed evidence from only part of the Chester canabae is considered in this volume, we can consider in outline how far the canabae as a whole might have fulfilled these roles. The social links implicit in role 1 are exemplified by the clearly shared cemetery at Infirmary Field and by the various relationships attested on tombstones: this topic will be considered in detail below.

It is natural to assume that roles 2 and 3 were the raison d'etre of all canabae, as of vici. The late first--early second-century pottery wasters found during the construction of the eastern sector of the Inner Ring Road in 1970 (see provisionally Swan 1984, 134 and fiche 1.239) provide an example of role 2. It is noteworthy that the forms produced -- mortaria and flagons -- were specialised, specifically Roman ones. Doubtless re-examination of the archives of excavations in the eastern sector of the canabae would reveal more examples of industry. However, as Casey (1982, 129--30) pointed out, the industrial and commercial life of vici and canabae in general is difficult to detect archaeologically, except for metalworking, and it is possible that much of the activity was small-scale. Some relevant fieldwork has been published since Casey’s article (see especially Cool’s 2002 summary of craft and industry at  York), but an overall synthesis is still awaited. 

Role 3 is exemplified by the Castle Street building identified as the mansio. The way in which the constructional history of this building closely mirrors that of the fortress (eg derelictionc 120--180) implies that this extramural building, at least, was actually constructed by the legion. If the opportunity arises to carry out a review of excavated sites in the eastern part of the canabae, it will be interesting to see how closely and extensively their histories match that of the fortress, with similar possible implications for the relationship between the two. Incidentally, the question arises as to where official travellers were accommodated during the period when the building was derelict: whether inside the fortress, in the postulated official enclosure overlooking the river, or billeted on the canabenses.

The obvious British example of role 4 is York, situated on the road from London to the northern frontier, with important visitors being Septimius Severus, Constantius Chlorus and Constantine. Chester is unlikely to have had such an important role in this respect, having an unfavourable position in terms of overland transport (Carrington 1985, 16--18). The Irish Sea can admittedly be seen as a logistical frontier, as were the Rhine and Danube (Jones, A H M 1964, 2, 844); a better, although more distant, parallel is the Black Sea coast, along which supplies may have been shipped to the eastern frontier via the port of Trebizond (Whittaker 1994, 100). However, Chester was not an inevitable port of call for ships from southern Britain en route to the northern frontier, being tucked away at the head of an estuary which debouched into a corner of Liverpool Bay. On the other hand, it could have served as a despatch point for supplies drawn from the north-west midlands to be forwarded by sea to the northern frontier -- which makes the apparent constriction of the road from the west gate of the fortress to the assumed position of the port all the more perplexing. However, as an aside, we should not forget the Weaver--Mersey route as a convenient way in which supplies from Cheshire could have been shipped north.

Roles 5, 6 and 7 beg not only the questions of how the army was supplied but also how deep were the changes wrought by the Roman military presence on local society and economy -- the questions which were the starting point for this essay -- and are considered in detail later in this essay.
 

The population  of the canabae: origins and roles
We are fortunate that at Chester there are numerous inscriptions to supplement the evidence of archaeology in helping us to understand the population of the canabae. However, as a background we must first understand the make-up of the legion.

Table 1 (based on Mann 1983 with obvious omissions from Wright & Richmond 1955 supplied) shows the pattern of recruitment and veteran settlement for the two legions based at Chester. The picture is clearest for the late first--early second centuries. A body of soldiers in Legion II Adiutrix, formerly sailors in the Ravenna fleet, came from Aprus in Thrace. Otherwise, as was normal in Britain, recruits came mainly from Gallia Narbonensis, Spain, Italy and the early colonies established on the Rhine and Danube frontiers. Few Italians were recruited into existing legions from the time of Hadrian. 

The epigraphic evidence that we have for recruitment into Legion XX VV in the later second and third centuries suggests at least some continuity with earlier patterns, High rank and consequent mobility explain the far-flung origins of Flavius Longus, a tribunus militum from Samosata on the upper Euphrates (RIB 1, 450) and of M Aurelius Alexander, a praefectus castrorum from Syria Osroene (RIB 1, 490). Specialisation as a magister balestari (sic) may explain recruitment of the soldier from Novaria in Italy (CIL 5, 6632 + add), while the man who died near Theveste in Africa (CIL 8, 2080) could have been drafted into the legion by Severus. RIB 1, 523 provides an example of recruitment from Thrace in the third century. However, we should also note that members of the legion died at Arles (CIL 12, 678) and Nimes (CIL 12, 3182), possibly suggesting origin in both cases. In addition, Henig (2004, nos 45 and 32) has re-assigned RIB 1, 524 and 492, commemorating men from near Turin and Merida respectively, to the third century, while RIB 1, 538, the tombstone of a man from Brescia, could again be third century (Henig 2004, no 54). Thus it does appear that there was still a continental presence in the legion in the third century, and not just in the corps of short-term officers. 

However, in the western provinces as a whole through the second century and especially in the third century soldiers were supposedly drawn increasingly from the provinces in which they served, especially from the families of veterans who had settled in the canabae. (For overviews, see Dobson & Mann 1973; Mann 1983; Le Bohec 2000, 75--82). Dobson and Mann assumed that this trend towards local recruitment would have applied in Britain, but, in the absence of much explicit epigraphic evidence, they argued largely by analogy with other provinces. They also conceded that recruitment to the legions in Britain was late (mid-second century onwards) and that the province provided few senior officers. Their conclusion on local recruitment to the army in Britain was attacked by Saller and Shaw (1984, especially 142--3; doubts also expressed by Birley 1979, 104), who pointed to the low percentage of soldiers in the province (and also on the Rhine frontier) who were commemorated by family members compared with the much higher level seen in provinces where local recruitment is certain (eg Africa, Spain, Pannonia; also now, it seems, pace Saller and Shaw, Upper Moesia: Wilkes 1999, 99--102). However, in reply, Mann (1985b) reasonably pointed to the lack of ‘epigraphic consciousness’ among the civilian population of Roman Britain, which meant that the families of local recruits were less likely to set up inscribed tombstones than those from stone-using provinces. Moreover, as will be seen below, the matter of family commemoration is a complex one. At Chester the name of the veteran L Ecimius Bellicianus Vitalis (RIB 1, 495) suggests that he was a British or Gaulish recruit, while in the third century Iulius Vitalis, a fabriciensis of the legion who died at Bath, is explicitly a Belgus (RIB 1, 156).

Overall, military tombstones dated to the third century (in fact probably to the first half of the century) number twelve as opposed, for example, to thirty-one for the late first--early second centuries and nine for the rest of the second century when, as we have seen, the resident garrison was at a low level for long periods. This is consistent with the hypothesis of increased recruitment of the ranks from a local population who lacked the epigraphic habit. (See Table 2).

Tombstones from Chester preserve only four examples of commemoration by the ‘wives’ of men who were stated to be serving or retired soldiers (or vice versa): RIB 1, 491, 505, 507, 526 (523 is a possibility). All are third century; one man was a centurion, one a beneficiarius legati, one an actarius and one a veteran. Only in one case is origin stated -- that of the miles gregarius, who came from Thrace. However, there are also three third-century tombstones where soldiers continued the earlier practice of being commemorated by anonymous heirs (RIB 1, 490, 492 and 522): the first two were men from Syria Osroene and Merida. The first half of the third century is the period when one would expect family-formation at Chester to have been at its strongest, given that the legion was fairly static in its home base, local recruitment was supposedly increasing and the legal impediments to soldiers’ marriage seem to have been removed. The picture that may be emerging is one where family commemoration became more common among better-off soldiers and veterans (see Saller & Shaw 1984, 141, fn 67), perhaps those of local origin, whereas recruits from abroad, who may not have had an extensive family with them, tended to be commemorated by unrelated heirs. 

The position of women vis a vis  the Roman army has recently been the subject of a thorough study by Phang (2001) and the following summary draws largely on her work. From the time of Augustus men up to and possibly including the rank of centurion were not allowed to marry during service, and any pre-existing marriages were probably dissolved on enlistment. (Contra on the position of centurions: Hoffmann 1995 and Hassall 1999). This regulation did not prevent soldiers taking female companions, but it did create legal problems for the women and to a lesser extent for the children of these unions; these problems were gradually ameliorated until the ban  was eventually revoked by Septimius Severus in 197 (Phang 2001, especially 16--19; also Campbell 1978). Because their ‘marriages’ were not recognised, women risked losing their dowries if their men abandoned them, which would seriously damage their chances of another marriage, while if their men died suddenly, they again risked losing their dowries and were not recognised as  intestate heirs. Their children  were illegitimate, which again originally debarred them from being intestate heirs (Phang 2001, especially 33--5, 203--4, 221--3, 306--7). However, soldiers could institute both their women and children as heirs by testament using the simplified ‘military will’, even if they were non-citizens (peregrines) or Junian Latins (informally freed slaves), while Hadrian eventually allowed the illegitimate children of soldiers into the ranks of intestate heirs  (Varon 1997; Phang 2001, 217--18, 319).  Soldiers could legitimise their marriages on discharge, although any children born during service remained illegitimate. Nevertheless, if their mothers were Roman citizens, these children would still be citizens. Legionaries seem not to have received the right of legal marriage (conubium) with peregrine women, with the privilege of citizenship  for the children of the marriage (civitas liberorurm), despite the fact that this was the case both for auxiliaries and Praetorians. The grant to auxiliaries of conubium, together with civitas for themselves and their children, was a way of extending citizenship: as ‘old’ Roman citizens legionaries were expected to marry citizen women (Phang 2001, 57--74, 121, 332; grant of conubium with peregrines and civitas liberorum to legionaries: contra Wells 1997).

In frontier areas initially almost the only female citizens would have been those who had accompanied soldiers, and perhaps dependent traders and the like, from other parts of the empire. Habitual ‘marriage’ by soldiers into a relatively scarce native population to compensate could have put a strain on the local marriage market and created tensions with the male population. In any case, citizen soldiers seem not to have regarded non-Romanised peregrine women as suitable partners, while native women in turn may not have been attracted to relationships hedged about with legal disabilities and involving the possibility of transfer to other parts of the empire, long separations and a greater-than-usual risk of sudden death for their men. The nomenclature of soldiers’ women on inscriptions is overwhelmingly Graeco-Roman, suggesting that they were indeed citizens or at least sufficiently romanised to fit in and possibly able to usurp citizenship (Phang 2001, 84, 138, 154--5, 191--5). Moreover, in the early principate frequent troop movements would have left little opportunity for the creation of permanent relationships with women. In these circumstances, soldiers’ emotional needs would have been met by their comradeship with their messmates (Roxan 1991). However, it is not to be concluded that the army was sexually inactive in this period. Leaving aside homosexuality and prostitution, female slaves were normal sexual partners of Roman soldiers. The high rates of soldiers’ commemoration by freedwomen (libertae) on the Danube frontier in the first century  suggests that many of these slaves were manumitted (Phang 2001, 193--5, 231--44). After a number of generations there would have been more locally born women (including descendants of enfranchised freedwomen) who had inherited citizenship and could therefore provide partners of the same legal status and similar cultural background for the soldiers of an increasingly static army, who themselves were increasingly likely to be locally born. 

Roxan (1991) and Phang (2001, 404--9) have shown that variations in the marriage rate of soldiers deduced from inscriptions were chronological rather than geographical, as assumed by Saller and Shaw, and that there was a general trend from commemoration by comrades and unnamed heirs in the first century to family commemoration in the third. This accords with the legal, demographic and military situations outlined above. It is not surprising to find that  only a  low percentage of soldiers were commemorated by family members on the Rhine frontier, where most of the military tombstones are first century, or in Britain, where large-scale troop redeployments continued through the second century. Nevertheless, soldiers always seem to have married less often and at a later age than civilians: the inevitable uncertainties of their careers, cultural exclusivity resulting in a tendency to marriage within the military community (endogamy), and a low birth rate restricted the number of available women in a way that was self-perpetuating, while hope of promotion and aspiration to a higher standard of living for families may have delayed marriage. The succession of pay rises granted by Severus, when he recognised soldiers’ marriages, Caracalla (212) and Maximinus (235) may have raised their standard of living and  made it easier for them to support families (Speidel, M A 1992, especially 88;  Phang 2001, 138, 179--90, 224--7). However, it needs to be remembered that this picture is largely based on inscriptions and  therefore cannot inform us about the lifestyles of soldiers who lacked the  epigraphic habit.

The number of relevant inscriptions at Chester is very small, but the hints they give are consistent with the general trend, in that in the first half of the third century the legion was fairly static in its home base, local recruitment was supposedly increasing and the legal impediments to soldiers’ marriage seem to have been removed. Nor must we overlook the purely archaeological evidence of the Infirmary Field cemetery. This shows that men and women of apparently similar status were being buried in the same area in the second half of the second century with ‘intrusive’ rites. Unfortunately it is not clear whether these burials are earlier or later than the abandonment of the Antonine Wall c 165 and the reoccupation of the Chester fortress in greater strength. However, they do suggest that women may have formed a numerically significant part of the community; they were not obviously of low status; and their material culture does not mark them out as peregrines. 

Whereas the wives of senior officers and probably centurions could live with their husbands within the fortress, the accepted view is that the partners of rank-and-file soldiers would have lived outside, even after marriage during service was legalised (Hoffmann 1995; Hassall 1999. Hassall  suggests that some surplus contubernia may have functioned as ‘married quarters’).

Table 3 lists those stones from Chester which attest either in words or images other roles as well as, or rather than, serving soldiers: veterans (7), wives (10 including the five mentioned above), parents (4), an adult son (1), children (9), siblings (10), doctors (2), patrons (3), freedmen (2+), slaves (9+). Unfortunately, apart from the doctors, in no case is an actual occupation stated, and it is possible that men named in RIB 1, 518--20, 528, 533, 536--7,  542, 552 and  559-61 were, in fact, soldiers but did not record the fact. Nor do we know much about the geographical origins of non-combatants: apart from the Geek doctors, we know only of men from Corduba, Berytus and Salapia (RIB 1, 518, 519 and 542 respectively), who are among the  possible soldiers; we do not know the origins of any of the women, but we may suspect that many of them came from the same communities as the menfolk. 

By contrast with the tombstones of soldiers, those which attest non-military personnel or veterans seem to have seen a flowering in the third century. Even if we discount those which also attest definite or suspected serving soldiers, twenty-five out of forty-two dated tombstones of this category have been assigned to that century. Two of these are explicitly memorials of veterans (RIB 1, 517, 526). Four commemorate married couples (RIB 1, 526, 543, 563; Wright & Richmond 1955, 117) and nine represent lone women (RIB 1, 562, 565, 568; Wright & Richmond 1955, 119--21, 123, 127, 160). All of these women could, of course, be relatives of soldiers, while the couples could be veterans and their wives; if so, the men had shed their former military persona. We should also note Wright & Richmond 1955, nos 118, 122 and 129, together with Henig 2004, no 75, which again depict non-military-looking men.

Again, the settled presence of the legion at Chester, leading to the growth of citizen-families, seems the most likely explanation for this flowering, outweighing the effects of any loss of epigraphic consciousness. Meyer (1990) has shown how in a civilian milieu the erection of tombstones by heirs or named commemorators was an expression of the personal obligations and advantages which came with Roman citizenship. Thus an upsurge in the number of such epitaphs in north African towns followed the block grant of citizenship to those communities. Legionaries and veterans were, of course, by definition citizens, and their descendants at least became so, but it is possible that some inhabitants of the canabae were only enfranchised when the Edict of Caracalla gave citizenship to all free inhabitants of the empire in 212, although Table 3 only contains three examples people with the nomen Aurelius (RIB 1, 491, 522 and Henig 2004, no 52). In addition, as seen above, it is possible that the canabae were raised to the status of a municipium in the Severan period: such a change would not have affected the legal status of the inhabitants, although it may have raised their self-esteem. Finally, this expenditure may have been made possible by the ‘trickling down’ of the pay rises given to the army by Severus, Caracalla and Maximinus.

The population  of the canabae: size
Attempting to work out the number of people who lived in the canabae is fraught with difficulty. The use of model life tables may take us part of the way towards an answer, although the applicability of such universal models to specific environments in the ancient world is now questioned. However, in order to construct a model, let us assume that:

• Life expectancy for the male populations from which soldiers were recruited  (and to which they returned after service) approximated to the figures in life table Model West Mortality Level 4 (Scheidel 1996a, 117--18). 

• There was  a male-biased population, as suggested by the ancient sources (although whether this indicates infant femicide or is merely a reflection in of the low cultural status of women is unclear) (Phang 2001, 154--5, 297--305). Life table Model West Level 2 has therefore been used for females. This age-distribution was believed by Bagnall and Frier to match the evidence found in census returns in Roman Egypt (1994, ch 4, especially table 4.3).

• The legion needed in the region of 264 recruits per year to maintain its nominal strength of c 4800 men, but discharges averaged 120 men per year. The steep decrement implied by these figures presumably reflects transfers and early discharge as well as mortality (Scheidel 1996a, 118--23). Soldiers enlisted at age 20 (Scheidel 1996a, 99--111).

• The proportion of ‘married’ soldiers increased over long period, rising from around 5% at age 22 to 55% at 47 (Phang 2001, 409, table 14,  based on epitaphs from Africa, the Danubian garrisons and the Praetorian Guard)

• Women ‘married’ about age 20, which seems to have been common across the empire, except in elite circles (Shaw 1987; 1994, especially 37). However, it is possible that  soldiers’ women ‘married’ later than the norm, while widows may have entered a second relationship with another soldier (Phang 2001, 175--8 ).

• Soldiers’ partners had enough children to produce a stable population. However, this assumption is also open to question: ancient figures suggest that fewer than two children per couple survived to maturity, so that ‘military communities’ may not have been self-sustaining (Phang 2001, 297--305, 337--43).

• Soldiers who ‘married’ during service stayed in the canabae after discharge 

• The resulting totals should be multiplied by a notional 1.5 to account for soldiers’ relatives, traders,  slaves, unmarried veterans etc

More fundamentally, as already mentioned, the applicability of model life tables to ancient populations has recently been challenged, by Scheidel (2001a; 2001b, 13--32) in a reversal of his own earlier position. Evidence from early modern populations suggests that infant mortality could have been lower than predicted, while that at mature ages could have been higher. Thus the number of births required to sustain a given adult male population may have been less than predicted by the models while the adult civilian population may have been lower. Moreover, geographical and temporal variations in the disease environment could have given rise to considerable differences in local age structures. 

Hopefully, however, for want of anything better, these tables may still be 'useful to think with' in giving some indication of the order of magnitude for the extramural population. They also put the supposed trend to recruitment castris into perspective: at the rates of marriage and reproduction envisaged the ‘military population’ could not have have sustained anything like a full-sized legion; rather it might have contributed about 38%, and the balance of recruits would have had to be drawn from surplus population over a wider area. The evidence for continuing recruitment from other provinces has already been reviewed. In Britain the most likely sources of citizen recruits are the veteran colonies of Colchester, Lincoln and Gloucester (see Birley 1979, 104--6), although no examples from these places are actually attested at Chester. Local sources would be the descendants of both legionary and auxiliary veterans who had settled in the vicinity, the latter exemplified by the Middlewich and Malpas diplomas (on which see below). If service in the army was not to lead to depopulation in the communities providing the recruits, there must have at least three surviving children in each family. For the possible scale of the requirement -- admittedly assuming recruitment from the rural population --  see Table 5. (For a brief study of this phenomenon in the Netherlands, see Bloemers 1989, 184--5). 

Recruitment to the army would have resulted in a large number of surplus females in these other supporting communities, some of whom will have chosen to move to the canabae, either as soldiers’ companions or dependents (Allason-Jones 1999, 48). In addition, given the general practice of men marrying later than women, widows would have been common. The lone women attested on the tombstones from Chester may be examples of the last two categories. 

How do these purely theoretical population figures accord with what is known archaeologically of the canabae?  There are two approaches, but both are hampered by the lack of extensive, controlled excavation in most of the relevant areas and even of detailed summaries of what is already known. The second is also hampered by our lack of understanding of how the relevant building types were used.

Let us assume that the eastern canabae extended c 450 m from the fortress and  c 150 m north and south of what is now Foregate Street (Mason 1987, 160--3; 2001, 181), ie covering an area of 135,000 m²/13.5 ha. Osborne (2004, 168) assumes  town/village population density in ancient Greece  to have been in the range of 100--250 people per ha. If applied to the eastern part of the Chester canabae this would give a population of 1350--3375. However, at best such parallels are merely indicative and at worst could be seriously misleading.

An ultimately sounder approach is to start from the few known buildings. These seem to have been predominantly, if not entirely, strip buildings, and to have corresponded roughly to the normal dimensions of this  type, c 10 m x <28 m (Mason 1987, 162; Perring 2002, 55). If  we assume side streets to the south of Foregate Street roughly as suggested  by Mason (2001, 181) and that all the frontages were built up, there could have been c 215 such buildings. However, the lack of large-scale excavation at Chester means that we do not know whether we are dealing with open-plan industrial sheds or houses with sophisticated internal divisions, let alone whether they had upper storeys. (For the range, see eg Perring 2002, 57, fig 11). It is thus almost impossible to generalise about the number of people they may have accommodated. 

Narroll (1962) has suggested a cross-cultural figure for population density of 10 m² of roofed space per person. However, as has been pointed out  by Morris (1991, 31), this density should be regarded as a minimum rather than a mode: in the classical Greek world houses seem to have allowed two to four times as much space per person, while Wallace-Hadrill has suggested a range of 35--9 m² per head for Pompeii and Herculaneum (1994, 99). The position at Chester is further complicated because, if the social composition was anything like that suggested above, we may be dealing  with a large proportion of ‘housefuls’ of mothers living separately from their male companions, children and widows rather than ‘normal’ households or families. If we apply double or quadruple Narroll’s figure to the total suggested ground-floor area of these buildings we get a  population of  1505--3010, ie 7--14 people per building. Thus at the moment it is hard to see how this area could have housed more than c 3000 people, ie less than a third of the estimated population of the canabae of 9900 (although it must be remembered that only two-thirds of this total has actually been calculated; the remaining third is a guess). Either,  then, this part of the canabae was much more extensive and/or densely built up than currently appears to be the case, or other large extramural suburbs remain to be discovered, or a significant number of soldiers’ companions lived inside the fortress, as suggested by Hassall, or there are significant errors in the model. Alternatively, the frequent absences from Chester of large detachments of the legion may have meant that this theoretical demographic total was never reached.

Models of military supply 
The supposedly sparse population of LPRIA Cheshire was alluded to at the beginning of this essay. As yet we do not know how sparse it was, and therefore how much difficulty there would have been in obtaining locally the agricultural surpluses needed to maintain the occupying forces logistically, although it is clear that there were deficiencies in the range and quantity of manufactured goods available. The increase in the number of rural settlements which has been detected  (see  below) and the creation of small towns can be seen as deliberate attempts to remedy shortages by supplementing and perhaps reorganising the existing means of production. The slowly diminishing economic dependency of Britain as a whole on other parts of the empire has been the subject of overviews by Fulford (especially 1989), who saw a much increased reliance by the British garrison on intra-provincial civilian resources from the reign of Hadrian. This is consistent with the broader trend towards local self-sufficiency in the north-western provinces as a whole detected by Reece (1981, 29--34) and with Whittaker’s model of the creation of frontiers stimulating the development of marginal areas (1994, 97). However, the focus of Fulford’s attention was the overall pattern of the distribution of goods and produce rather than the landscapes of production or the precise mechanisms of distribution, although, citing Higham and Jones’ work in Cumbria, he minimised the reliance of army units on their immediate hinterlands. In the more detailed study of the Chester canabae and the hinterland of the fortress these detailed  considerations are important, and, as mentioned above, military occupation does seem to have produced quantitative and qualitative change in the occupation of Cheshire.

The Chester resource area
The Roman army’s demand for raw materials, manufactured goods, foodstuffs and animals was huge and wide-ranging (in general, see Breeze 1984). Although we cannot estimate how much local production increased or what level it reached,  we can make some attempt to quantify the needs of the army.

Focussing on foodstuffs, the aspect which has received most attention has been the demand for wheat. Usually this calculation has been carried out on a large scale, with the objective of estimating the burden represented by provincial armies (see, for example, the useful review of studies of the British situation by M E Jones (1996, 204--16)). More detailed estimates were made of the impact of forts in Cumbria on the local economy by Higham and Jones (1985, 107--9). Such archaeological evidence as there is for wheat consumption at Chester has not yet been assembled.

Meat was also part of the military diet. The main food animals consumed by the Chester garrison were cattle, pig and sheep/goat in that order, although because of differential destruction of the bones the exact relative frequency is difficult to establish (Cartledge 1991 unpublished), let alone absolute quantities. This pattern is consistent with that found at other legionary sites in Britain and is ultimately derived from Gaul and Germany, although the high percentage of pig may also reflect the Mediterranean view of pork as a high-status food (King 1984, 1999a, 178--83; 1999b). The importance of pig-rearing in LPRIA Cheshire is unknown. Sheep may have been kept mainly for milk and wool rather than as a source of meat. The frequency of the different species changed over time, clearly reflecting further economic and social changes, but it is not intended to explore this topic here.

The consumption of cheese and other dairy products by the Roman army in Britain is also known, as is that of vegetables, salt and Mediterranean foodstuffs, especially those transported in amphorae -- olive oil, wine and fish sauce -- but again the quantities are unknown. Cheese production was recognised at the farmstead at Birch Heath near Tarporley (Fairburn et al 2002, 79--80, 107--8) and could have been associated with cattle- or sheep-rearing.  (For an overall review of the Roman military diet, see Davies, R W 1971; for further examples of foodstuffs see Bowman 1994, 64--74).

In order to assess more precisely the impact of some of these demands, we need to try to quantify them and to model the pattern of local rural settlements, their population and productivity. Following the pattern that seems to have prevailed in Cumbria, arable areas of 6 and 12 ha surrounding dispersed farmsteads have been used for the model: Higham and Jones cite field systems of 6--10 ha at Wampool on the Solway estuary (1985, 72) and in excess of 16 ha at Severals on Crosby Garrett Fell (1985, 103--4). Beyond these field systems may have been meadows (50 ha?) for growing winter fodder for livestock, and beyond that rough summer grazing. There is as yet no archaeological evidence from Cheshire for these areas, although it does exist for the type of settlement. A settlement density of one site per 2 km² has been recorded on north Merseyside (see below).

It is worth observing that, by Mediterranean standards, field systems in the region of 12 ha (48 iugera) would have been large: allotments of only 2--10 iugera (0.5--2.5 ha) were common (Salmon 1969, 178, nn 110--11), although around Pola and Bologna areas of 50 iugera (12.5 ha) have been found (Garnsey 1998, 111, 127). The smaller of these Italian allotments were undoubtedly cultivated intensively as market gardens; without additional labour the largest allotments, in the region of 12 ha, seem likely to have been used for the extensive farming of wheat and for stock-raising.

Given that many of the rural enclosures of Roman Cheshire only contained a single roundhouse (see below), let us assume that each farm was worked by a single family, albeit one that may have extended at times to three generations, with help from exchange labour. The age structure of the family used in the model has been determined as follows:

• Life expectancy for men followed mortality table Model West  Level 4 (the same as for soldiers); that for women Level 3 (adopted by Saller 1994, 23 as a general-purpose table for the ‘ordinary population of the empire). These figures are subject to the caveats set out above; in particular, the rural disease environment(-s) would have been different from the quasi-urban one of the  fortress and canabae.

• Age at marriage is assumed to be 20 for women. However, not only did marriage  practices vary from one province to another, they also varied between town and country, and it is from the former that our largely epigraphic evidence for a wide gap between men and women in age at first marriage generally comes. It is possible that in the country both women and especially men married a few years earlier (Scheidel 2005b, 12--14; 2006, 6). Age at marriage for men is therefore set tentatively at 22.

• Given that many of the rural enclosures of Roman Cheshire only contained a single roundhouse (see below), let us assume that each farm was worked by a single family, albeit one that may have extended at times to three generations, with help from exchange labour. 

• Grandparents would have joined their son’s families, unmarried aunts their nephew’s. The model deliberately maximises the size of the family, which in practice would have changed through the life-cycle of a generation. In addition, as already noted, the mortality tables used may exaggerate the chances of grandparents surviving.

However, Huntley (in Huntley & Stallibrass 1995, 57--9) reported that barley rather than wheat was the most commonly found cereal on Roman sites in north-west England, spelt wheat coming second. Small quantities of emmer wheat, bread wheat and oats were also found. The same seems to have been true of north Wales (Davies, J L 1997, 268--9).  Traditionally barley was used by the Romans as animal feed and for punishment rations, but it is possible that its prevalence in western Britain indicates that it came to be adopted as a normal part of  the human diet, military and civilian (see Huntley 2000, 351--5 on Ribchester) -- as indeed it was in Attica. Certainly, higher yields might be expected from cultivating barley, but these would have been at least partly offset by greater weight losses during threshing and a lower calorific value than wheat (Whitby 1998, 114--18).

For productivity levels we are again forced back to documentary evidence. From Columella and from calculations concerning the late third-century tax system of capitatio-iugatio, Duncan-Jones (1982, 49--50; 1990, 204--5) estimated that one man could work the equivalent of 3 ha of grain. Spurr (1986, 137--40) gives more detailed figures without seriously affecting the conclusion.  Admittedly  these figures are for Italian  agriculture, but White’s comparanda from early post-medieval England and north America (1970, 413--14) suggest that they are relevant.  On this basis it seems unlikely that more than 6 of a total 12 ha could have been  put down to grain, even by larger families, the rest being used for other crops, including vegetables for domestic use or for animal feed, or left fallow. Likewise, smaller families may simply not have been able to cultivate more than half of 6-ha arable farms for cereal. Again we must expect the area under cultivation to have fluctuated over the household life cycle. (For the possible roles of women in agriculture, see Scheidel 1995, 1996b).

Following Hopkins (1980, 119; 2002, 199), one option is to assume a tax/requisition rate of 10% of gross production. On the other hand, higher rates are quoted by Duncan-Jones (1990, 189--93), including 33% tax-cum-rent on imperial estates in Africa Proconsularis. We therefore present alternative tax/requisition rates of 10% and 30%. From any theoretical surplus we must also allow for threshing- and storage losses.

The following observations may be made on the calculations set out in Table 6:

• 6-ha farms could only rely on being self-sufficient in wheat if the yield was above c 0.5 t per ha

• 12-ha farms could only rely on being self-sufficient in wheat with a yield of c 0.5 t per ha if the tax rate/requisition rate was c 10%

• Both 6-ha and 12-ha farms could support a 30% levy if yields were c  1 t per ha

In fact, in Cheshire the quantitative importance of cereal production in Roman times is currently unknown. One would expect it initially to have been inadequate to produce the surpluses needed and to have been encouraged on purely logistical grounds (cf Groenmann-van Waateringe 1989 on the lower Rhine frontier). As will be seen, the pollen evidence  does seem to show an increase through the LPRIA and Roman periods; the form of local rural settlements also suggests an increase in arable farming. On the other hand, the hints of meat-curing and leather production at Nantwich show that cattle-rearing continued to be important locally, as it was in the LPRIA (see below). 

Turning now to domestic animals, few attempts seem to have been made to calculate the quantities consumed by the army. J L Davies (1997, 271) has suggested c  450 g per man per day, Groenmann-van Waateringe (1997) as little as a sextans, c 55 g.  Both figures seem extreme, but it has seemed worthwhile extrapolating from Davies’ figures, basing the calculations on cattle. If we assume 184.5 kg of meat per carcass, then a full-strength legion would have consumed 4273 beasts per year, necessitating a total breeding herd of 17093. If we again assume that one hectare of meadow land could have produced winter fodder for one animal (Higham & Jones 1985, 106), then it seems likely that dispersed settlements on a 2 km² spacing as envisaged could each have supported herds of  about fifty  head of cattle. In fact, if we are correct in envisaging family homesteads practising mixed farming as the characteristic element in local rural settlement, then the size of herds may have been constrained by the availability of time to manage them rather than the area of land. If we assume a herd of twenty cattle, providing one carcass per year for domestic consumption and four for tax or sale, then 1068 herds would have required, spread over c 2136 km^2,  equivalent to a circle with a radius of 26 km. On the other hand, small herds of only eight head of cattle, providing only one animal per year for tax or sale, would have occupied over 8500 km² (radius 52 km), while large ones of fifty head would have occupied 684 km^2, (radius 15 km). In any case, it is reasonable to assume that cattle would have been driven to Chester on the hoof and slaughtered on the outskirts of the settlement.

However, we need to remember that in fact pigs were more commonly eaten at Chester than cattle, and that many of the former may have been kept on the fringes of the settlement. The area of  land required should therefore probably be reduced

Arguably as important in the economy as food production would have been secondary animal products, especially wool and leather,  both of which tend to be neglected in general discussions because of their rare survival. Although the south-eastern half of England was prime country for sheep, they would nevertheless probably have been ubiquitous in Roman Britain. Wool was sold combed or in the fleece; the time-consuming task of spinning was a major task for the female population and was centred on the home, but weaving does not seem to have been so gender-specific (Wild 2002). 

Wild has suggested that the army in Britain may have been supplied with three basic pieces of clothing (tunic, cloak and blanket) in alternate years. On the basis of experiment, each may have taken six person-weeks to weave and would have required the wool from three to four sheep. On the other hand, papyrological evidence suggests a shorter time, in the order of 3.3 weeks per garment.  In addition, it may have taken five handspinners to supply one weaver. A variety of modes of production can be envisaged: individual household production (c 11.5--21 weeks per garment including spinning and weaving, assuming a mother and  daughter split the spinning equally); cooperative household industry (eg c 13.5--24 weeks per head for a team of three women and their daughters, producing a total of four garments: three to meet the needs of one adult in each household and one for sale/tax); and individual or nucleated workshops with full-time weavers relying on ‘putting out’ the spinning. The technology immediately suggests the need for social labour, whether co-operative or hierarchical. Wild suggested that much of the textile sold in Britain may have been produced on villas estates using seasonally available surplus labour; the role of towns in textile production is unclear. The only significant increase in production for sale/tax  beyond that reached by ‘cooperative household industry’ seems to be achievable by reducing the number of garments taken by the producers or increasing the number of weeks per year devoted to the task. Given the contribution to agricultural tasks which they were undoubtedly required to make, it seems unlikely that women on many of the small rural settlements such as are found in Cheshire could have done more than satisfy the needs of their own households for clothing  unless they cooperated with their neighbours. If, for the sake of example, we assume  cooperative household industry  as a mode of production, the 7200 basic pieces of clothing needed each year by a legion would have required the work of 21,600 households -- far more than were needed to ensure a food supply.

The military demand for leather -- for tents, shield-covers, shoes, belts, saddles and other horse gear -- would also have been large and could only be met after the Roman introduction of vegetable tanning (van Driel-Murray 2002). The largest demand, for tent leather, would have probably been occasional, and in Britain often seems to have come from cattle, rather than the goatskins used on the continent (Hooley 2002, 339--40; goatskins at Ribchester: Howard-Davis 2000, 307). In frontier areas  van Driel-Murray argued that specialised leather goods were produced at the legionary fortress, with only shoes being made and repairs carried out at auxiliary forts. The evidence for the repair and reuse of leather objects from Vindolanda suggests that even at the start of the second century supplies were still scarce; they did not become plentiful until the Hadrianic--Antonine period.

It was observed above that the frequent movements of legionary detachments made it difficult to calculate the population of the canabae. Likewise we can attempt to calculate the food requirements of the legion at full strength, but again that would have been a rare occurrence, and we can only guess how far any growth in local supplies would have been geared specifically to Chester rather than to the garrisons of north-western Britain in general. One might look for signs of agricultural specialisation after the manner predicted in Von Thünen’s Isolierte Staat, with horticulture and dairying close to the fortress and canabae, an emphasis on arable further out, and ranching beyond that (see Morley 1996, especially 58--63,  for an application of Von Thünen’s model to Rome and Italy. 

Supply mechanisms
The pattern in the first and early second centuries seems to have been for individual soldiers to be supplied with many of their needs -- clothing, boots and socks, weapons, tents, hay-money and food -- by the army, for which stoppages were made of about 40% of basic pay. Further deductions were made towards the camp Saturnalia and ad signa. Perhaps as early as Hadrian’s reign these stoppages seem to have been reduced, with soldiers having to buy supplies for themselves, either through the army or directly from traders  (Speidel, M A 1992, 93--4, 97--8). During the second half of the second century emperors are thought to have begun the provision of free supplies (annona) which became a feature of the later third century in particular as the value of regular pay declined through inflation and as troop movements in response to invasion required emergency provisioning (Rickman 1971, 278--83; Casey 1974, 48--50). Nevertheless, the third century also witnessed an increase in cash donatives to soldiers, which actually became more important than the allowances in kind (Duncan-Jones 1990, 105--17), and traders continued to supply soldiers to the extent that their exaggerated prices were the ostensible reason for Diocletian’s edict on maximum prices in 301 (quoted by Breeze 1984, 279).

Much of the available evidence for the organisation of military supplies comes from Egyptian papyri, but there seems no obvious reason why the picture they paint should not apply to Britain. Much remains obscure, and it is likely that the reality was complicated (for a recent review of some of the issues, see Adams 2001). Nominally at least, they were the responsibility of the provincial procurator, and the basic supplies needed for each unit would have been agreed between his officium and that of the unit commanders (Davies, R W 1971, 123; Mitchell 1976, 124--5; Breeze 1984, 281; Remesal Rodriguez 1990). Grain at least was then acquired -- whether bought (frumentum emptum) or gathered as tax in kind -- on the procurator’s behalf for individual units (Rickman 1971, 271--2). Further supplies of food and other goods could be bought compulsorily at a fixed price by units with the procurator’s permission (Breeze 1984, 277--8). However, the Vindolanda tablets of the late first--early second centuries show the extensive financial independence retained at that time by local commanders and the wide range of goods bought ad hoc by soldiers from traders, presumably in the vicus (Onken 2003, 193--200, who also notes that the rarity of mentions of grain in the Vindolanda tablets suggests that it reached the fort through a redistributive mechanism). 

The collection of army supplies, whether purchased or gathered as tax in kind, might be organised directly by soldiers (eg Bowman 1994, 40; Breeze 1984, 281; Adams 1999) or under contract by negotiatiores, who could at the same time trade on their own account, bringing in additional goods and supplies which were available for purchase by individual soldiers and by civilians in the canabae (Whittaker 1989, 69--72; 1994, 104--13). As already mentioned, some of these goods and supplies could have been gathered as tax but sold on as surplus to government requirements. The actual transport, at least overland, of official supplies and officials within a province could be requisitioned from communities en route (Mitchell 1976, passim; Black 1995a, 5--6, 8, 11, 22) and was indeed an obligation on communities supplying requisitioned produce (Mann 1985a 22, discussing Tacitus Agricola 19). However, it is unclear how far it would always have been practical for transport to be provided by a dispersed rural population, and this may again have been provided  by contractors (cf  Brunt 1990,  379 and 531 on the gathering and delivery to a central point of rent-grain by the conductores of imperial domains in north Africa, rather than by the individual cultivators). 

The logistics of army supply overlap with those of taxation in general and the exercise of local administration. Taxes, of course, again came within the remit of the provincial procurator. In urbanised areas (eg among the British civitates, including the civitas Cornoviorum), responsibility for the gathering and delivery of direct taxes lay with the city authorities, although they might again employ contractors for the purpose (Brunt 1990, 531). However, in other areas, including the ‘military zone’ of northern Britain, a greater organising role would probably have been played by the army (MacMullen 1963, 60, 68--9; cf Adams 1999 on Egypt). Black (1995a, 63) has suggested mansiones as delivery points for taxes, as well as being likely places for the exercise of local government functions by army officers (cf MacMullen 1963, 60; Wilson 2002, 456). Here we should note the presence of a commissariat officer (annonarius) at Vindolanda who was also possibly a centurio regionarius (Tab Vindol 2, 255; Bowman 1994, 132--3, no 28). (On centuriones regionarii, see below). As already noted, the land between the mansio at Chester and the southern fortress defences was apparently open space (Mason 1980, 85). Although there were sound military reasons for maintaining a clear zone outside defences, as observed in the western extramural area at Chester, this more extensive space north of the mansio could have been used not only as paddocks for mounts and baggage animals but also perhaps for penning requisitioned livestock.

G D B Jones (1984, 79) also suggested mansiones as the sites of markets, citing the similarity of their layout to, for example, the primary forum at Sabratha, the Caesareum at Cyrene and the Caesarean market at Corinth. In his more recent and detailed study, Black has shown that mansiones were often preceded by deliberately founded roadside vici and were accompanied by tabernae to meet needs not met by the mansio itself and to cater for travellers who were not entitled to use the cursus publicus. In addition, of course, they required their own support staffs (1995a, 3, 4, and especially 10--11 on the Pizos decree). Thus it can be seen how they may have gained a function as local market centres even if not primarily intended as such. Finally, as Shaw (1981, 58) has argued, periodic markets provided convenient occasions for  gathering taxes from a dispersed population. Thus, in sum, we may perhaps envisage mansiones not only as providing lodgings for official travellers but also sites where markets could be held under supervision, to which taxes and army supplies could be delivered, and other government functions exercised.

Wherever possible, goods would have been transported over long distances by water. Middleton (1979, 95) suggested that shipping of supplies around the coasts of Britain was the responsibility of the Classis Britannica or other units such as the Numerus Barcariorum attested at Halton on Lune near Lancaster (RIB 1, 601), although the latter may have been concerned exclusively with lighterage around Morecambe Bay (inf Dr D C A Shotter).  In addition, Fulford (2002, 93--6) has argued that Legion II Augusta had an important role in the shipping of supplies along the west coast, which might explain the distribution of black-burnished 1 pottery. He also pointed out that, in addition to the Classis Britannica, Legion VI had its own helmsmen (RIB 1, 653) and drew attention to the Chester tombstone of an optio who died in a shipwreck (RIB 1, 544). However, in addition to  the traders operating between both the Rhineland and Bordeaux and eastern Britain is M Verecundius Diogenes, attested  as a sevir Augustalis at York and presumably a civilian, who may have been a moritex (shipper) (Hassall 1978, 43). While the military may well, therefore, have had a direct role in the shipping of supplies, this need not have been to the exclusion of others, who may have operated under contract to the military or freelance. Moreover, we must distinguish between responsibility for the shipping of supplies, which may have been in the hands of the military, and that for assembling the cargoes, which could have been left to negotiatores.

From the west coast three people have been suggested as merchants: the anonymous man who dedicated an altar at Bowness-on-Solway asking for divine support for his venture (RIB 1, 2059); on the basis of his eastern name, M Antigonus Papias of Carlisle (RIB 1, 955) (both cited by Breeze 1984, 283); and T Aurelius Ateco of Old Carlisle (RIB 1, 902: Davies, R W 1977). To these we may possibly add, again on the basis of their Greek names and apparent absence of military connections, Callimorphus and Thesaeus from Chester (RIB 1, 558). 

We may therefore envisage that local supplies and taxes, when not delivered directly to Chester, may have been gathered from the local population at mansiones and perhaps other central places, from where they were then forwarded in bulk either by the military or by contractors under military supervision. If Chester gained municipal status in the third century it is possible that  greater responsibility was assumed by civilian authorities, as seems to have happened in Egypt; presumably the magistrates of the civitas Cornoviorum also had some responsibility for the delivery of supplies from further south. Negotiatores are likely to have played a larger role in the supply of  staples and especially manufactured goods from further afield, whether from southern Britain or the continent. 

The modes of production and acquisition of supplies adopted by the army seem to relate to the level of development of the economies of frontier provinces. Together with the establishment of military works depots such as that at Holt (floruit c 90--120/130), the compulsory deductions from soldiers’ pay in the first century suggest a rigid system occasioned by weaknesses both in production and distribution in the civilian economy of the province. By contrast the changes in the Hadrianic period, such as the closure of the works depots and the widespread use of black-burnished ware (on which see for example Gillam 1973, especially 54)  and possibly changes in footwear (van Driel-Murray 1985, 55--8), indicate a more fluid situation brought about by greater civilian productive capacity.

Settlement in west and mid-Cheshire and adjoining areas
The environs of Chester
About two km to the south of the Chester fortress was the settlement at Heronbridge referred to above (Mason 1988a, 1988b, 174--9, 2002d, 2003a, 2003b). This appears to have consisted of ribbon development along Watling Street (here represented by the modern Eaton Road), although its full extent is still unknown. It seems to have originated as a small-scale industrial and agricultural settlement cAD 90 and consisted of strip buildings. One oven may have been used for corn-drying. In the Hadrianic period it was rebuilt in more substantial form, with signs of overall planning, the buildings being arranged in groups of three in insulae perhaps 120 Roman feet square (ie 1 actus). Many of these buildings had glazed windows, opus signinum floors and stone eavesdrip gutters. A column capital and an altar dedicated to the Matres Ollototae could suggest the presence of a temple. Some buildings may have come to the end of their lives towards the end of the second century and there was some rebuilding in the early third century, but occupation of the settlement as a whole only decreased markedly late in the third century, although several new buildings, one with a hypocaust, were constructed in the fourth century. Detailed reconsideration of the pottery is needed to refine this chronology. Funerary banquet-style tombstones, probably third-century, found during recent excavations are a further indication of prosperity. Cremation burials have been found lining the road south of the settlement, while the tombstones mentioned above were associated with interments on the river bank near its centre (cf the comments above on riverside burials at Chester).

The settlements on the continent analogous to Heronbridge have been listed by Mason (1988a, 138). Some of these were founded in regions where there had been pre-Roman nucleated settlements, eg Carnuntum, Aquincum, Singidunum, Vindobona (Mason 1988b, 178), while others, eg Oppidum Batavorum (Nijmegen), Vetera, along with Oppidum Ubiorum (Cologne), were artificially created in regions without a pre-Roman urban tradition (Carroll 2003, 28), although there may have been a LPRIA cult centre at Nijmegen (van Enckevort & Thijssen 2003, 61). It may be specifically these Rhineland sites that Dio Cassius had in mind when he wrote (lvi, 18 on the early Romanisation of Germany) of ‘cities being founded: the natives ... were becoming accustomed to holding markets and were meeting in peaceful assemblies. Hence, ... under careful watching ... they were becoming different without knowing it.’ Some attracted a local population, others immigrants (eg on the Danube: local population -- Aquincum; immigrants -- Carnuntum, Viminacium (Mocsy 1974, 141); on the Rhine: local population -- Vetera, Novaesium (Mason 1988b, 177); immigrants -- Oppidum Batavorum (van Enckevort & Thijssen 2003). 

Mason (1988a, 148; 1988b, 177--8) following Mocsy (eg 1974, 140) has emphasised the probable differences in legal and administrative status -- especially the ability to own land, whereas in the canabae it could only be rented -- in explaining the existence of the canabae and the Heronbridge-type settlements so close together. While these differences may have acted as a stimulus to the growth of the Heronbridge-type settlements, perhaps the aim, at Chester as on the lower Rhine, was to create a proto-urban nucleus in a region of dispersed native settlement. If this was the intention at Heronbridge, then the attempt either failed or was spectacularly successful: the material culture of the site is so like that of the fortress and canabae as to be indistinguishable and no native influence has yet been traced. 

It is possible that the reconstruction in the Hadrianic period was related to that emperor’s attempts to tighten discipline by separating military from civilian (SHA Vita Hadriani, 10), in this case by enhancing a local centre for those for whom the fortress was an attraction but whose business did not require them to live in the canabae: for example traders who wished to do business with local civilians as well as the military; possibly veterans, who are attested at the comparable settlement at Great Bulmore near Caerleon (Vyner 1978). Mason (2001, 150) has suggested that the Julius Secundinus who dedicated the altar to the Matres Ollototae at Heronbridge could have been a veteran of the ala Hispanorum Vettonum civium Romanorum, which was stationed in Wales in the first century AD and then transferred to Binchester, where dedications to the Matres Ollototae are also found (RIB 1, 1030--2). The regular plots associated with the Hadrianic rebuilding imply a unified development and could suggest official involvement, now if not before, in its layout and its construction. However, they could also simply show the influence of a body of vicani familiar with Roman building regulations (Perring 2002, 56; Sommer 1991, 475; 1999, 87--9), with the groups of buildings being financed by wealthy individuals. However, so far houses of more sophisticated plan are absent.

Matthews (2002, 408--11) has drawn attention to the existence of a pattern of rectangular fields south of Chester cut diagonally by the Chester--Wrexham road (A 483), which is assumed to follow the line of the Roman road from Chester to Caer Gai, and favours the view that it represents pre-Roman landscape clearance and organisation. However, this suggestion ignores the fact that this field system is aligned on Eaton Road/Watling Street and should therefore be Roman if not later. Moreover, the Chester--Caer Gai road may actually have run through Heronbridge. The system does not appear to have the standardised measurements characteristic of true centuriation, although detailed survey has not been carried out, and it is possible that it is actually medieval and belongs to the lost village of Claverton, which lay in this area. (There is possible evidence for true centuriation east of Manchester: Richardson 1983, 1986, 2004).

Excavations in the autumn of 2003 on an extension to the Chester Business Park 3.5 km south-south-west of the city centre have revealed a further site which lay near the western edge of the putative field system. This seems to have consisted of timber structures surrounded by cobbled surfaces and wells. The earliest structures were round, those in later phases rectangular; surrounding them were small plots, enclosures and droveways. The site seems to have been used for stock-penning rather than human habitation and it may be significant it lies on the western edge of the Eccleston ridge, where the latter runs down to the Saltney marshes, and would thus have been adjacent to heavier land better suited to pasture. The finds, presumably scatter from an occupation site, consisted of the usual range of pottery found at Chester, extending from the early second to the mid-third century, plus ceramic building materials, a little metal and glass. . It is possible that the site operated as an outlier of Heronbridge practising dairy farming. Alternatively, it may have provided a holding area for cattle being driven to Chester from further afield for slaughter.

Further to the north, again on the edge of the Saltney marshes, was another settlement, which was in existence from the late first century probably into the fourth but with its floruit in the late second. The buildings seem to have been simpler than those at Heronbridge, apparently consisting of stone-founded, rectangular houses with tile roofs, perhaps set in ditched enclosures. Nevertheless, a wide range of pottery was in use, including samian and fine wares; other finds included an intaglio, shale amulet, querns and iron tools (Newstead 1935).

Small towns and industrial settlements
Thirty km to the east of Chester lay the industrial settlement at Wilderspool at the junction of the Chester--Manchester road along the south bank of the Mersey (Margary 1955, route 701) and King Street through mid-Cheshire and west Lancashire (Margary 1955, route 70a). The settlement had a planned street grid, which suggests official control over its early development, its main axis being formed by King Street. A fort has been conjectured on the north bank of the Mersey but not yet found, although military fittings show the presence of soldiers. Seven phases of occupation have been detected, extending from the end of the first to the early fourth century. The earliest buildings were timber with tiled roofs. There is little evidence for their function but their inhabitants were probably meeting military contracts. In the second to the fourth centuries stone buildings with columns and capitals were to be found in the central zone. It took time for the rearward areas of the settlement to be built up, but they were being used for industrial purposes by phase 3 (140--70). Although a variety of metalworking was carried out here, the most recognisable product was pottery, which was distributed throughout Cheshire and, in the case of mortaria, up to the Antonine Wall. This industry declined sharply after the 160s, presumably because of a change in military contracts after the withdrawal from the Scotland, and some of the potters seem to have moved north to Carlisle; coincidentally there was a sharp decline in coin loss on the site (see below). Nevertheless, the late second--early third century saw intense industrial development, perhaps serving the local civilian market rather than the military. More substantial buildings were erected in phase 5 (220--50) with stone footings. In phase 6 (250--70) industrial activities were being carried on in former domestic areas but still in substantial buildings. Phase 7 (270--320) was purely agricultural. A stone courtyard building in the Loushers Lane area with associated round houses standing in enclosures has been interpreted as a mansio or villa; it may have succeeded a LPRIA farmstead. 

A little about the population of Wilderspool can be inferred from the types of pottery made there and the names of the potters attested on the stamped mortaria, although these names have not yet been the subject of specialised study. Although pottery production was probably under way in the Flavian period, that of red-slipped ‘raetian’ mortaria started c 110. The style of these vessels, and that of other some types of red-painted pottery, implies a link with Rhaetia and Upper Germany, most probably through the army. Among the potters’ names attested on mortarium stamps in the Hadrianic--Antonine period, Brico is presumably Celtic, Ovidius Roman. CCM and DIS/LDB may be abbreviations of the tria nomina carried by Roman citizens. Nanieco seems to have worked previously at Hartshill-Mancetter in the midlands. (For a summary of the site see Strickland 1995, especially 28--46; for the pottery industry, see Hartley & Webster 1973; Hartley 1981; Swan 1984, 104--5. The above structural account is based on Rogers et al forthcoming).

Over the past few years much excavation has also been carried out at the industrial sites at Middlewich and Nantwich. At Middlewich a fort was occupied through the Flavian--Trajanic period, with saltworks located in the vicus to the south. The discharge diploma granted in AD 105 to an unknown trooper in the Ala Classiana civium Romanorum, found at Middlewich (RIB 2 (1), 2401.3) may show the continuing attraction of life in the vicus to veterans. The saltworks survived the closure of the fort, and indeed the accompanying settlement expanded in the Antonine period, its alignment shifting to that of King Street to the east. The buildings were large examples of the normal ‘strip’ type (Garner 2004--5; Dodd 2004--5). Other industrial activities supposedly included tanning, iron-smithing, a little pottery-making, copper- and leather-working, cobbling and weaving, although no detailed evidence has yet been published (Petch 1987, 2002--8; Strickland 2001). On the basis of coin loss, Shotter (1998--9b, 57) has suggested an upsurge in activity in the first half of the third century, possibly connected with increased activity at Chester, followed by a sharp downturn in the second half of the century (see also below). Local soil conditions make the third--fourth-century stratigraphy difficult to understand. However, there is pottery of the period (Garner 2004--5) and the saltworks continued in production into the fourth century, when a lead salt-pan attests the name Lutamus or Clutamus (Petch 1987, 222--8; Tomlin & Hassall eds 2005, 486).

At Nantwich, excavations in 2002 produced evidence of salt-production from the late first to the early third century (Connelly & Power 2004--5). The first phase of activity began in Hadrianic times and continued through the rest of the second century, consisting of two large timber-lined brine cisterns, two roadside buildings (one a barn or salt store, burnt down at the end of the second century). In the early third century the two buildings were replaced by paddock and a corral in the centre of the site. The brine cisterns were backfilled with numerous finds and replaced by wicker-lined pits associated with wooden troughs. The paddock and corral suggest that the salt-making process was extended to include meat-curing and leather production. The proportion of samian in the pottery assemblage could indicate a military hand in the operation and it is argued that a fort may yet be found in Nantwich. A military or at least official interest does seem likely given the scale of the operation, but there is as yet no direct evidence. There are also two lead salt-pans, probably fourth century, which attest a Cunitus cler[icus] (RIB 2 (2), 2416.2--.3), while there are also three from Shavington nearby, two of which attest a Flavius Viventius episcopus (Penney & Shotter 1998, 2000--01). 

Salt extraction is also attested at Northwich, where there were Flavian and Hadrianic forts, with pottery and iron furnaces also operating during the hiatus (Petch 1987, 199, 224). Five lead salt-pans have been found there, one stamped Veluvius and another conceivably EVE[ … (RIB 2 (2), 2416.6; Petch 1987, 201; Penney & Shotter 1998, 363--4). Jones and Mattingly (1990, 224--6) also record salt production just over the Shropshire border at Whitchurch. 

The nature of the Roman salt industry in Cheshire is not yet properly understood. Increasingly it is looking as if we are dealing with polyfocal production, with sites scattered across the south-eastern half of the county rather than a centralised extraction industry with a single centre, as appears to have been the case at Droitwich (Fielding & Nevell 2004--5, 66--8). This militates against the industry being run as a single, directly operated imperial estate, although one leased out to private contractors is possible, and, as noted above, the scale of operations at Nantwich certainly suggests governmental interest. Here we should note that salt extraction was already well established in Cheshire by the Roman period, spanning the whole of the Iron Age. At the moment only one Iron Age production site is known for certain, at Railway Farm, Moston, near Crewe, but others are probable, eg at Middlewich, where there are hints of  high-status Iron age occupation (Nevell 2004--5, 12--13; Garner 2004--5, 23--4) or possible, eg at Nantwich, where dendrochronological dates could suggest pre-Roman activity (Reid et al 2004). By the fourth century parts of the industry had apparently fallen into the hands of the church, but it is unclear whether we can extrapolate anything from this about the earlier ownership of the industry (Shotter 2004--5). 

The exact purpose of the settlement at Nantwich is difficult to understand at the moment. The evidence for major activity under Hadrian (now strengthened by the discoveries reported by Reid et al 2004) suggests a connection with the northern frontier. Clearly productive capacity was being increased: as noted above, in the earliest years of the second century leather was still in short supply in the north and did not become plentiful until Hadrian’s time and later. The key questions surround the animals from which the hides came. Were the latter by-products of domestic meat consumption by the local civilian population, bought or commandeered for the army? On the other hand, if the carcasses were destined for the army as well, then they too would presumably have been butchered and  cured at Nantwich for distribution elsewhere. If so,  a close connection with Chester seems unlikely: it would be easier to drive cattle to Chester on the hoof and then slaughter them and tan the hides at the destination.

The distinctive vessels made of VCP (Very Coarse Pottery) in which Cheshire salt was transported in the Iron Age and the earliest years of the Roman occupation allow us to track its distribution, north to the Ribble, south through the west midlands to Gloucestershire and south-east to Derbyshire, Nottinghamshire and Leicestershire (Nevell 2004--5). No such distinctive vessels have yet been recognised later in the Roman period, so the distribution and type of sites which the salt reached at this time is unknown. Undoubtedly much of the production was destined to satisfy military requirements, but, given the Iron Age background, an extensive civilian market also seems likely.

Like the Chester canabae and possibly Heronbridge, these settlements also had their rural fringes.  Field ditches have now been  found in the environs of Middlewich and Nantwich, the former enclosures used for pottery production and saltworking as well as agriculture (Philpott 2005, 11, citing M Leah pers comm).

Also within the purview of the Chester garrison, and certainly under direct government control at some time, were the silver/lead mines of Flintshire, in the te