York's ‘African-style’ Severan Pottery Reconsidered

ABSTRACT This article examines the phenomenon of so-called North African-style pottery made in early third-century York. The pottery, which was produced in significant quantities in late Ebor ware, is strikingly different from vessels in circulation in Roman Britain and the north-west provinces – so much so that the late Vivien Swan argued that it was ‘made by Africans for the use of Africans’. The present study reassesses the evidence of ceramic genealogical influences, production waste, fabric supply, consumption patterns and contextual finds associations. The results shed new light on the manufacture and use of late Ebor ware by York's military community, qualifying claims made about the repertoire's links with novel culinary practices, cultural diversity and the unique historical circumstances of Severan York.


R
ecent interest in a globalised and multicultural Roman Empire has brought fresh attention to archaeological evidence for cultural diversity. 1In this discussion, so-called 'North African-style' 2 pottery in Roman Britain has become a prominent case study. 3Since the cultural significance of the pottery is debated, the present study examines the evidence in more detail, focusing on York, which features the largest published assemblage of such wares in Britain.The most authoritative studies on the subject remain those of the late Vivien Swan, who developed detailed arguments on the connections between British and North African pottery repertoires. 4Swan's work focuses on pottery production and typological influences in tandem with historically plausible movements of Roman military personnel.To move the debate forward, the present study explores what late Ebor ware did in Severan York, as 1 On Roman cultural diversity, see Eckardt 2010.On Roman globalisation, see Pitts and Versluys 2015; Witcher  2017.   2  informed by the concept of 'objectscapes'.An objectscape refers to the material and stylistic characteristics of a repertoire of objects in a given period and geographical range, emphasising its connectivity, relationality and impact. 5Exploring objectscapes entails confronting the following questions.What objects, styles and materialities are new?Where do they come from?How do manufacturers innovate?What are the historical consequences of these material changes?Facing these questions, I review the genealogy of York's late Ebor ceramic repertoire, considering it in the context of local pottery production and supply, its spatial distribution and consumption, and contextual associations with other pottery and archaeological finds.From this perspective I aim to shed new light on the broader significance of the pottery, rather than fixating on its capacity to serve as proxy evidence for North African migrants. 6RK'S 'NORTH AFRICAN-STYLE' POTTERY: THE DEBATE SO FAR The possibility of North African influences in locally made oxidised and grey fabrics termed 'Eboracum wares' at York was first recognised by J.R. Perrin on examination of pottery excavated at 37 Bishophill Senior in 1973. 7Perrin noted a distinctive series of deep bowls or 'casseroles', often with sagging bases, that were typologically reminiscent of vessels from Roman North Africa. 8The vessels had 'no real antecedents in the ceramic repertoire of Roman York, and scarcely any in Roman Britain as a whole', and Mediterranean pottery specialist J.W. Hayes considered them to be local imitations of North African forms. 9Casseroles in heat-resistant oxidised fabrics are typical of Mediterranean cooking practices, but not those of the north-west provinces. 10Finds of such vessels in York and the northern frontier zone raise the possibility of transfers of Mediterranean cooking practices, otherwise scarcely observed in Britannia.The corpus of casseroles from York grew into the 1990s, with new finds recorded in later publications by Perrin and J. Monaghan. 11Monaghan's corpus of pottery from an extensive range of published and unpublished sites of Roman York includes a comprehensive re-evaluation of the North African-style vessels, resolving uncertainties over the dating of the late Ebor repertoire and placing its first production from c. A.D. 200. 12 This study confirmed the North African-style vessels to be part of a repertoire of novel forms never before seen in pottery assemblages in the city, with short-lived production dating to the first three decades of the third century.
Alongside the work of Perrin and Monaghan, the North African connections of late Ebor ware were further examined by Swan.Swan's work brought to light finds of casseroles listed in older publications and other late Ebor types that seemed to show North African connections, including dishes and lids (FIG.1). 13Considering the quantities of pots with North African genealogies made in early third-century York, Swan argued that the phenomenon could only be explained by the presence of customers who desired the pots for their specific culinary functions, being 'made by Africans for the use of Africans' or by other soldiers from adjacent Mediterranean regions. 14Experimental work revealed that vessels with sagging bases could be used to bake bread when used with 5 Pitts and Versluys 2021.The term draws attention to changes in material culture, in much the same way that 'landscape' and 'cityscape' address changes in land use and urban topography.

6
This aim follows the contributions in Van Oyen and Pitts 2017, which strive to move beyond the 'representational' use of objects in historical research to unlock the broader potential of material culture.
reconstructed North African-style ceramic braziers. 15Other pottery evidence from York hinting at North African connections was re-evaluated, including a series of distinctive head-pots made in Ebor-like fabrics and convincingly dated to the Severan period (FIG.2; Monaghan form YH2). 16 Finds of over 50 head-pots from York, which share plausible resemblances to the portraiture of Severus' Syrian wife Julia Domna and his son Caracalla, are best understood as a FIG. 1.Late Ebor ware North African-style vessels from Roman York (illustrations, types and find numbers after Perrin 1981; 1990; Monaghan 1993; 1997).
consequence of the visit of the imperial family in A.D. 208-11, and coincide with the floruit in production of North African-style cooking vessels.
In later publications Swan made the case for other locally made vessels from Severan York to have close typological links with other parts of the Roman world, namely the Rhineland and Gallia Narbonensis. 17Furthermore, Swan and Ray McBride's analysis of Ebor-ware kiln dumps revealed a small number of rarer vessels with distinct typological connections to the Rhineland, including a series of idiosyncratic mortaria. 18The collective evidence raised the possibility that North African potters in York were accompanied by at least one potter from the Rhineland and perhaps also one from Gallia Narbonensis.Focusing elsewhere in northern Britain, Swan published further studies highlighting the appearance of equivalent North African-style wares, with the strongest evidence from Carlisle and the Antonine Wall. 19fter a hiatus in study, York's North African-style pottery received fresh attention in discussions on Roman human mobility and cultural diversity.Most notable is Michael Fulford's review of the evidence for pottery and migration in Roman Britain. 20Fulford accepted the possibility that North African-style pottery in York could have been made by migrant potters from Africa Proconsularis (Tunisia), on the strength of its typological connections. 21However, he also highlighted significant problems of linking pottery finds to the consumption habits of cultural or ethnic groups.Fulford pointed out the following weaknesses of Swan's thesis: (1) that little is known of Roman military consumption habits in North Africa to provide reliable comparisons; (2) that North African pottery production was heavily influenced by wider Mediterranean traditions from the late Hellenistic era, raising the possibility that the influences impacting late Ebor ware could have come from elsewhere in the Mediterranean; (3) that the absence of ceramic braziers in York casts doubt on the direct transfer of culinary practices from Africa Proconsularis; and (4) that fresh quantitative analysis was needed to assess better the popularity and consumption patterns of the new vessels.
The substantial issues raised by Fulford underline the complexities of determining the genealogies of pottery vessels in an interconnected Roman world, let alone linking configurations of objects with the cultural identities of mobile people.For these reasons, the present consensus is cautious about Swan's interpretation of the late Ebor repertoire. 22Monaghan accepted the likelihood of a North African influence in production, but was doubtful of a direct link with North African consumers, believing Roman pottery manufacture to be largely producer-led, with the implication that many users of the late Ebor repertoire in York would have been local. 23Likewise, Michel Bonifay included York in a survey of North African-influenced pottery in the Roman Empire without fully endorsing Swan's case for pottery 'made by Africans for the use of Africans'. 24A major reason why Swan's hypothesis remains attractive concerns the relationship between changing pottery production and contemporary developments in Severan York.This evidence is summarised in the next section, before returning to the pottery in detail thereafter.

SEVERAN YORK: A NEW CITYSCAPE AND OBJECTSCAPE
To appreciate the significance of late Ebor ware, we must consider the historical and archaeological context from which it emerged.Historically, the fortunes of York at this time are framed by the activities of legio VI Victrix, which garrisoned York's fortress from c. A.D. 122, and the visit of the emperor Septimius Severus in 208, who used York as a base for major military campaigns further north, before his death in the city in February 211. 25 Legio VI had previously supported the claim of the rebel Clodius Albinus, who was defeated by Severus at the Battle of Lugdunum in A.D. 197.The remnants of legio VI are thought to have returned to their base in York later in A.D. 197 or shortly thereafter, no doubt strictly reprimanded for siding against Severus and probably reinforced by officers and soldiers loyal to the new regime. 26This is one possible scenario for the transferral of soldiers of North African origin to York. 27Another constitutes the visit of Severus to York in A.D. 208, when he was likely accompanied by legionary vexillations (possibly from legio III, based in Lambaesis in North Africa) for his expedition to northern Britain. 28Aside from the emperor, who was born in Tripolitania, definitive proof of the presence of North Africans in York at this time is lacking, although there is corroborative epigraphic evidence from the northern frontier zone. 29One example is the tombstone of Gaius Cossutius Saturninus from Birdoswald, a member of legio VI from Hippo Regius (Africa Proconsularis). 30A large number of Roman skeletons from York have been revealed to have probable North African origins, as determined through the application of stable isotope and osteological analyses. 31It is possible, if not certain, that some of these are of Severan date.York was made capital of Britannia Inferior by A.D. 213 in the reign of Severus' son Caracalla (who also visited the city in A.D. 208-11) and was probably promoted to the status of colonia around this time or shortly after (by A.D. 237). 32rom an imperial perspective, the Severan visit for a brief time made York the political and military centre of Britannia, if not the whole Roman world.Aside from its strategic location at the edge of the northern frontier zone, the city was a good candidate to host the imperial entourage, having undergone rapid and far-reaching urban development in the late Antonine period. 33Excavations on the site of St Leonard's Hospital in 2002-04 suggest that the stone refurbishment of York's fortress defences, including the addition of a series of imposing projecting towers to the south-western wall, was complete by the late second century. 34lsewhere in the fortress, the evidence for structural change at this time is limited. 35A possible North African influence in the modification of the fortress baths is suggested by the find of an earthenware tubulus lingatus at 4-5 Church Street (FIG.3, site 6), in a Roman sewer that was out of use by the early third century. 36The same kind of tube had widespread contemporary use in North Africa and is known in Britain only at military bases undergoing Severan refurbishment, most notably the legionary fortresses at Caerleon and Chester. 37However, this suggestion remains unproven, since the finds lack direct architectural associations and the proposed North African origin of the tradition may simply reflect the better state of preservation of baths in North Africa compared with those elsewhere.
Outside the fortress north-east of the Ouse, the so-called canabae became the site of significant productive activity in the early third century. 38This was the location of a series of dumps of late Ebor ware at 21-33 Aldwark and Peasholme Green (FIG.3, sites 2 and 3), suggesting the presence of nearby kilns. 39The same site at Aldwark produced evidence of broken coin moulds for issues of Severus and Caracalla, leading to the interpretation that they may have been used as officially sanctioned forgeries to satisfy a surge in local demand at the time of the emperor's visit in A.D. 208-11. 40 Further productive activity in the canabae is attested at 16-22 Coppergate (FIG.3, 27 Birley 1966, 728; Swan 1992, 6; Monaghan 1997, 842.28 Birley 1988, 175; Swan 1992, 6; 2002, 62.   29 Swan 1992, 3-4; 2008, 71-2; Eckardt 2014, 68-71; Nesbitt 2016, 234-5.30   RIB 3.3445; Ottaway 2013, 302.31 Leach et al. 2009; Montgomery et al. 2010; Müldner et al. 2011.32 Ottaway 2004, 83; Ottaway et al. 2020, 72.33 Ottaway 2013, 213-16; Ottaway et al. 2020, 195.34 Ottaway 2013, 155; cf.Ottaway 2004, 71-5.35   Monaghan 1997, 865. 36 Nesbitt 2016, 235; cf.Sumpter 1976, 45, no.23; Mason 1990, 221.On the date of the sewer assemblage, see Monaghan 1997, 1059.37 Wilson 1992, 118-19; Nesbitt 2016, 235.38   For use of the modern term canabae, see Royal Commission on Historical Monuments (England) 1962, 18;  Ottaway 2004, 87.39 Monaghan 1997, 873-4; Swan and McBride 2002.40 Magilton 1986, 40; Cool 2002, 9; Ottaway 2004, 89.Based on recent examination of the moulds, Richard Brickstock (pers.comm.)suggests that they could have been made later in the third century, while not completely site 7), with finds of glass-making equipment revealing the largest manufacturing site found to date in Roman Britain. 41The Ebor ware glass-melting pots are typologically dated to the late second and early third centuries, lending weight to the suggestion that the enterprise was connected with the Severan visit. 42While there is room for caution in ascribing such loosely dated archaeological developments to Severus' visit in A.D. 208-11, the importance and duration of the occasion could certainly have warranted the unprecedented scale of the productive activities that took place to the east of the fortress, supporting the notion that the canabae was under direct military or state control at the time. 43cross the Ouse in the civilian settlement known to archaeologists as the colonia, the late second to early third century ushered in a period of major growth and investment. 44The latest accounts on the development of Roman York south-west of the Ouse underline its rapid development and the appearance of monumental buildings from c. A.D. 160-70, suggesting a FIG. 3. Schematic plan of Roman York, with sites mentioned in the text: 1. St Maurice 's 1972; 2. 21-33 Aldwark;  3. Peasholme Green; 4. York Minster; 5. 9 Blake Street; 6. 4-5 Church Street; 7. 16-22 Coppergate; 8. Wellington  Row; 9. 5 Rougier Street; 10. 24-30 Tanner Row; 11. 1-9 Micklegate; 12. 37 Bishophill Senior; 13.Clementhorpe  Nunnery.
ruling out a Severan date.He also concedes that, although some officials may have turned a blind eye to the forgeries, they were unlikely to have been officially sanctioned.
41 Cool et al. 1999; Cool 2002.42 Cool et al. 1999, 150, 158.43 Ottaway 2004, 87.44   The term colonia is often used in archaeological literature on York (and the present study) to refer to the Roman period civilian settlement south-west of the Ouse, but strictly speaking it probably applied to the whole city from the early third century.
deliberately planted community that was most populous and prosperous in the Severan era. 45The late second to early third century witnessed major construction phases at Wellington Row, 5 Rougier Street and 24-30 Tanner Row (FIG.3, sites 8, 9 and 10), although the functions of the new buildings are uncertain. 46To the south, a thriving centre is suggested by the construction of a substantial bath-house at 1-9 Micklegate (FIG.3, site 11) and the movement of thousands of tons of earth and rubbish to create a residential terraced area spanning c. 200-300 m at 37 Bishophill Senior (FIG.3, site 12). 47Outside the assumed defences of the colonia to the south-east on the site of a later Benedictine nunnery at Clementhorpe, further evidence of early third-century terracing suggests the extension of the affluent residential area at Bishophill, with the construction of the first phase of a high-status building (FIG.3, site 13). 48These developments underline a number of far-reaching changes that had come to fruition by the Severan era, affecting most major functional and administrative zones of the city.

YORK'S LATE EBOR REPERTOIRE: GENEALOGIES AND PRODUCTION
Returning to the pottery evidence, this section examines the typological closeness of material from York and the Mediterranean, expanding the comparisons drawn by Fulford between late Ebor ware and equivalent vessels in Bonifay's North African-type series. 49Taking a complementary approach, I examine the plausibility of parallels for a representative range of supposed North African-style vessel types from Monaghan's 1997 corpus, privileging comparanda in circulation in the early third century. 50tarting with the casseroles, FIG. 4 compares vessels in the late Ebor repertoire with typologically similar types from publications on Mediterranean pottery. 51The illustrated vessels reveal not only wide variation in the morphology of late Ebor casseroles (Monaghan form BA, showing five of the six vessel subtypes), 52 but also significant typological likeness to vessels of North African manufacture.While the likenesses between individual vessels vary, complete examples such as 2978 show striking similarities with their North African counterparts, in this instance comprising the vessel's exterior profile and its sagging and grooved base. 53ignificantly, the North African vessels with the closest similarities to the late Ebor repertoire in FIG. 4 come from sites and cemeteries on the Tunisian coastline, with Bonifay's suggested dates for each ranging from the late second into the third century. 54This information underlines the likelihood of geographically specific links between York and Africa Proconsularis, spanning multiple contemporary vessel types and coinciding with the dates of major changes in York's military garrison.
associated with ascribing the York types with a specific North African genealogy. 57 5 are from Tunisian coastal sites, where the vessels in question circulated at the start of the third century. 59Likewise, the closest Mediterranean prototypes for the late Ebor lids (FIG.5, Monaghan form LD), Hayes types 185 and 196, either share chronologies spanning the late second to third century or have specific Severan dates, 60 with all the examples again coming from Tunisian coastal sites.Despite evident typological similarities, it is important to retain caution in ascribing North African genealogy to the late Ebor dishes and lids.This is because the typological distance from other dishes and lids in use in the north-west provinces is not sufficiently large to rule out the possibility of local or regional influence, unlike the casseroles, which lack equivalent predecessors.In particular, some of the late Ebor ware dishes look very similar to relatively common types in circulation in the lower Rhineland, which also overlap chronologically and could conceivably have influenced the York potters. 61he strength and consistency of North African typological traits in late Ebor ware casseroles, supplemented by the appearance of dishes and lids of plausible North African genealogy, underline the likelihood of a Severan connection between York and Tunisia.This is preferred to the alternative scenario of indirect and generalised inspiration from wider Mediterranean potting traditions, which would be expected to impact the north-west provinces in a much slower and diffuse manner.The distances involved are too great to explain such close reproduction of North African genealogical elements in local production at York, at least without some accompanying movement of potters. 62A direct link is unsurprising given what is known of the exceptional historical circumstances of Severan York, with North African military personnel possibly present in the reconstituted legio VI, and moreover in vexillations of legio III taking part in Severus' northern campaigns.In addition to their shared genealogies and chronology, another factor linking the vessels in question seems to be functionalcasseroles, dishes and lids are all associated with cooking and eating fooda perspective that is examined further below.
Considering the whole late Ebor repertoire, it is worth briefly considering the genealogies of vessels lacking North African characteristics to appreciate better contemporary production and consumption patterns.These vessel forms comprise butt-shaped jars, u-profile bowls and flagons (FIG.6, Monaghan forms JB, BU, FE and FT).Swan and McBride make the case for northern Gallic and Rhenish typological affinities for the butt-shaped jars (JB) and u-profile 57 Hayes 1972, 49-50, 200-2, types 27, 181.The types illustrated by Hayes come from locations as dispersed as Tharros (Sardinia), Knossos, Tarragona, Corfu, Sabratha and Rome (type 27), and Corfu, Tunisia and Rome (type 181).

58
Monaghan 1997, 1016.The same rings are present in most examples in Bonifay 2004, 215, fig.114; cf.Swan  1999, 412-17.These vessels have markedly different profiles from Pompeian red ware dishes from York, for example Monaghan 1997, 884-5.59   Some of the York forms show affinity with earlier variants dating to the mid-second century (Bonifay 2004, culinaire B type 5, var.A), which is unsurprising given the slow evolution of the form.The domed lid (Bonifay 2004, culinaire C type 11, var.A).

61
For example, dish Stuart 10, which appears in funerary contexts in the Netherlands and Germany in the second and early third centuries (Stuart 1963, 26-7; Haalebos 1990, 143).

62
The situation is reminiscent of the appearance of distinct North African visual schemes within the mosaic pavements at Rudston Villa, located less than 40 km to the east of York.The schemes are similarly without parallel in Roman Britain, and likely reflect the desires of a patron with North African experience (Cousins 2017).bowls (BU). 63They suggest that the vessels are paralleled in a series of everted and curved lid-seated jars from the Rhine-Moselle region, 64 with the u-profile bowls being produced in some of the same workshops. 65On closer inspection, these genealogical links are less clear-cut than for the North African-style vessels.Concerning the butt-shaped jars, only selected examples of type JB2 with lid-seating (for example FIG. 6, no.59) have obvious links with examples from northern Gaul.The strongest parallels consist of late second-to third-century vessels from cemeteries in the Tongeren region.These vessels share equivalent traits such as analogous rims for lid-seating, a pair of parallel shoulder grooves and similar vessel FIG. 5. Typological comparisons of late Ebor dishes (PA) and lids (LD) with North African vessels (illustrations, types and find numbers after Hayes 1972; Perrin 1981; 1990; Monaghan 1993; 1997; Bonifay 2004).6).Swan and McBride identify a pinched-neck flagon from Peasholme Green closely paralleled at Soller (Germany), but this type is rather different from Monaghan's type FT, which exhibits general affinities with trefoil flagons produced across the north-west provinces and beyond.While no specific parallels have been advanced for Monaghan's type FE, it nonetheless bears a remarkably close resemblance to Gose type 382, of the first half of the third century, 68 falling into the group of late Ebor forms with more or less prominent Rhenish influences.
A final vessel of relevance to the long-distance ancestry of late Ebor ware is the so-called Dales or Dales-type jar, which appeared in York in a range of local fabrics (including Ebor ware) at roughly the same time as the major changes in the Ebor ware repertoire (FIG.7, Monaghan type JD). 69The type is recognised by its distinctive triangular lid-seated rim, which is paralleled in vessels from Languedoc, Gallia Narbonensis, from the late second to early third century. 70It should be noted that the parallel involves rim profiles alone, since the complete vessels look rather different.Swan believed the wheel-thrown type (JD2) to have been introduced to the region by migrant potters, forming the prototype for the handmade Dales-shelly ware jar (JD1) produced elsewhere in Yorkshire. 71aken together, the eclectic genealogies of the pottery types making up the late Ebor repertoire are remarkable in mid-Roman Britain, implying a high intensity of mobile people and styles, as embodied in York's rapidly globalising objectscape.However short-lived, equivalent wholesale stylistic and morphological changes to local pottery production are unusual in Britain after the watershed shifts in pottery repertoires that accompanied the Iron Age to Roman transition. 72To explore this further, the rest of this section expands the discussion by considering the production of late Ebor ware in York.
Although no kilns have been found, production evidence is strongest in the early third century, consisting of large quantities of kiln waste found at 21-33 Aldwark and Peasholme Green (FIG.3, sites 2 and 3), seemingly redeposited after the main phase of production had ended, c.A.D. 225. 73The early third-century production of late Ebor vessel forms coincided with a decline in the fabric diversity of Ebor ware, with white-slipped, red-painted and Ebor 3 variants all terminating.Most typical of the Severan era are the long-lived wares Ebor 1, a 66 Tongeren examples include graves 7 and 31 (Vanvinckenroye 1963), Pannenovenstraat grave 3 (Vanvinckenroye 1970), graves 74, 76, 78, 121 and 277 (Vanvinckenroye 1984).Further examples, from the rural cemetery of Wanzoul, c. 35 km to the south-west of Tongeren, include graves 60, 80, 99, 136 and 247 (Destexhe 1989)  Monaghan 1997, 873-4.This concords with material from earlier excavations at the nearby Borthwick Institute (King 1974), which was incompletely recorded and is unsuitable for quantitative analysis.coarse oxidised orange-red fabric, and Ebor 2, a coarser variation distinguished by the addition of extra sand. 74he main late Ebor pottery types in the material from the kiln dumps at Peasholme Green and 21-33 Aldwark quantified by Monaghan are presented in TABLE 1.These data permit some cautious insights into production priorities.At Peasholme Green, the most common types are casseroles (BA, over 18 per cent) and lids (LD, 26 per cent), with North African-style vessels FIG. 6.Other vessels in the late Ebor repertoire (illustrations, types and find numbers after Wenham 1968; Perrin 1981;  Monaghan 1993; 1997).
making up roughly half of the assemblage.While a broad range of equivalent vessels are present in the Aldwark dumps, these are dominated by u-profile bowls that share typological affinities with material from Colchester and the Rhineland.The flagon forms FE and FT are present in small quantities, mirroring their thin distribution across the city.The discrepancies between the assemblages from Peasholme Green and Aldwark serve to underline the diversity in late Ebor ware production.Rather than presenting a homogeneous repertoire, the evidence raises several possibilities.One is finer-grained phases of production, with the Aldwark material possibly marking the final phase of Ebor ware manufacture. 75Another is the operation of multiple potters each responsible for the production of stylistically different suites of vessels or the existence of sub-markets of consumers with preferences for different vessel types.2. 77 Unfortunately, quantified data on the samian ware present in excavations from 5 Rougier Street and 24-30 Tanner Row are missing from the published report and microfiche. 78For this reason, to visualise patterning in pottery supply to Severan York, two sets of correspondence analysis (hereafter CA) are presented: the first dealing with a larger sample of sites excluding the samian ware (FIG.8) and the second including the samian for those sites for which it is reliably quantified (FIG.9).Site locations are shown in FIG. 3, and colour-coded in CA according to major zones of the city.CA is a well-established method for the comparison of Roman artefactual assemblages. 79The results of these analyses are explained below, with For this broad-brush comparison, it should be borne in mind that assemblages with longer date ranges may feature residual material.While residual fabrics have been excluded from the comparisons, it is not possible to exclude all residual forms for longer-lived fabrics still current in the early third century.See Monaghan 1997 for further discussions on residuality for these assemblages.77 Data are selected from Perrin 1981; 1990; Monaghan 1993; 1997.Exceptionally, material from 37 Bishophill Senior is included in TABLE 2, having been originally quantified by rim numbers.In CA, this material has been incorporated in FIG. 8 as supplementary data, meaning that it is compared indirectly, without affecting the disposition of material quantified by EVE.78 Perrin 1990.79 See, for example, the pioneering work of Cool et al. 1995, 1626-47, on finds from York; Cool and Baxter 1999.For applications to pottery data, see Perring and Pitts 2013; Pitts 2019.80   Data are presented as percentages for the ease of visual comparison in TABLES 2 and 3, but raw quantities are used in CA in FIGS 8-10.Considering York's Severan fabric supply as presented in FIG. 8, several clusters of sites and wares emerge, allowing major differences in pottery provision to be summarised in terms of the functional zones of the city.Ebor ware correlates most strongly with kiln dumps in the canabae, as well as fortress assemblages from 9 Blake Street (upper-right quadrant).A loose cluster of assemblages from civilian and residential sites in the southern zone of the colonia corresponds with Rhenish and Nene Valley colour-coated wares, black-burnished ware 2 and grey-burnished wares (upper-left quadrant).The other assemblages, mainly from the colonia near the riverfront, seem to reflect a secondary market for Ebor ware, in addition to other largely coarse fabrics (lower half of FIG. 8).
The patterns seen in FIG. 8 are reinforced in FIG. 9, this time including samian ware.Samian is most prevalent at Wellington Row, followed by 1-9 Micklegate and the residential sites to the south in the colonia, where it is plotted alongside other fineware and imports (lower-right quadrant).Civilian areas south-west of the Ouse stand out as major consumers of imported and fineware pottery in the early third century, although it is possible that the high figures from Wellington Row represent dumped stock from a riverside warehouse. 81Otherwise, the most striking clustering in FIG. 9 is the sharp distinction between assemblages from the fortress and canabae on the left and assemblages from residential and civilian sites in the colonia to the The sharp observable distinction in the pottery supply of military and civilian sites in Severan York is of no small significance for understanding the emergence of the late Ebor ware repertoire.While Ebor ware is present in quantity in all assemblages considered (including varying amounts of residual material), supply generally drops off further away from production areas in the canabae, followed by the fortress.Only assemblages from 24-30 Tanner Row rival the military areas of York for their relative proportions of Ebor ware, although it should be noted that the percentage figures in TABLE 2 for Tanner Row are artificially high owing to the absence of quantified samian. 82Taken together, these data lend support to the notion that Ebor ware was produced primarily on military-controlled land, but not exclusively, for military consumers.In contrast, the communities of the flourishing civilian parts of York seemingly had a greater predilection (and means) to obtain higher-quality fineware and imported ceramics.It is important to be cautious about the social implications of these patterns.The weight of evidence suggests high levels of interaction between the military and civilian populations of fortress FIG. 9. Correspondence analysis: fabric supply in late second-and early third-century York, quantified by EVE and including samian ware.

82
McComish 2015b, however, suggests military connections at 24-30 Tanner Row based on its meat supply, stamped tiles of legio VI and the probable reuse of timbers from the fortress.towns like York, involving women, children and men engaged in a variety of civilian activities. 83rom this perspective, it is possible to envisage a scenario in which the late Ebor repertoire served as 'barracks ware' for preparing food by groups of soldiers on active service.If soldiers and other members of the military community had social connections across the river in the colonia, in those contexts it was probably more appropriate to use suites of pottery better suited to civilian life.This idea is further explored in the following section on the contextual associations of late Ebor ware vessels.
Before leaving the subject of fabric supply, it is worth briefly discussing the complementary evidence of amphorae.York's amphorae assemblage is notable for its substantial quantities of North African fabrics and vessels, which are more prevalent here than anywhere else in Roman Britain, at least based on data available in the late 1990s. 84Although it is not possible to date most of the material with more accuracy than ascribing it to the third and fourth centuries, an early third-century surge would be unsurprising given the substantial needs of the Severan military expedition. 85Quantified data presented in TABLE 3 single out the fortress site at York Minster as the most voracious consumer of North African amphorae, 86 including 11 of 127 sherds from an alley deposit between barrack blocks that is dated to the early third century. 87hile moderate quantities of North African amphorae at canabae sites underline a likely link with military supply, York stands apart from the rest of Britannia in this regard, where equivalent finds are more commonly associated with non-military establishments. 88

THE LATE EBOR REPERTOIRE: SPATIAL DISTRIBUTION AND CONTEXTUAL ASSOCIATIONS
To get a sense of the spatial distribution of different vessel types in the late Ebor repertoire, TABLE 4 summarises the numbers of vessels from excavations for which reliable quantification has been undertaken.Material from sites previously discussed is supplemented by assemblages from George Street and Dixon Lane, 89 situated just across the Foss in the canabae, and the cemetery at Trentholme Drive, located to the south of the colonia. 90Apart from sites close to production areas in the canabae, the distribution of types is uneven.The full repertoire of North African-style types including head-pots is best attested within the fortress at 9 Blake Street and York Minster, with moderate presences in the canabae at George Street and Dixon Lane, and in the colonia at 24-30 Tanner Row.Larger numbers of casseroles from 37 Bishophill Senior are paralleled by equivalent vessels in older excavations nearby, 91 and are suggested to come from redeposited fortress waste used in the construction of the substantial early third-century artificial terrace. 92In contrast, vessel types argued by Swan to have Rhenish ancestry (JB and BU) are scarcer in fortress assemblages than at 37 Bishophill Senior.The jar forms JB and JD are especially popular in the cemetery at Trentholme Drive, seemingly following similar biographical pathways to equivalent vessels placed in graves in northern Gaul.Where direct associations with skeletal remains are possible, JB vessels were found with a pair of males aged 25-40 and JD vessels were found with a pair of adult females aged over 40; but the 83 For example, Allison 2013, 319-43; Haynes 2013, 130-3.84 Williams 1997, 967.85 Williams and Carreras 1995, 237-8.86   Data are from Williams 1990; 1995; 1997.87 Williams and Carreras 1995, 237.88 Williams and Carreras 1995, 237.89   McComish 2015a.Pottery data courtesy of Nienke Van Doorn, York Archaeological Trust.90 Gillam 1968.91 Ramm 1976.numbers are too small to draw conclusions. 93Elsewhere, the late Ebor repertoire is incompletely represented at most sites lacking military associations.Recent excavations at Heslington East in the hinterland to the south-west of York likewise produced few Ebor ware vessels with North African genealogy. 94o further scrutinise the data summarised in    Roskams and Neal 2020, 89.visual interpretation, the assemblage patterns (FIG.10a) are separated from the distributions of vessel forms (FIG.10b), which are distinguished by fabric where relevant.A list of codes used in this analysis is provided in APPENDIX 1. 95 Considering the disposition of pottery assemblages in FIG.10a, the patterning is consistent with the analysis of fabric supply in FIGS 8 and 9, with the main distinction being retained between assemblages from the fortress (FIG.10a, centre-left), canabae (FIG.10a, lower-right) and colonia and civilian areas (FIG.10a, upper half).The main difference from the preceding analysis concerns the clustering of assemblages from 37 Bishophill Senior with those from the fortress.This association concords with Monaghan's hypothesis that the Bishophill assemblages include redeposited fortress waste.In view of this interpretation, the Bishophill assemblages are coloured blue in FIG.10a to denote a fortress connection.
The corresponding pottery forms in FIG.10b confirm the predilection for imported and fine colour-coated drinking vessels in the colonia (upper-right quadrant) and hint at greater samian use at colonia sites closer to the Ouse, most notably 5 Rougier Street, 24-30 Tanner Row and Wellington Row.In contrast, assemblages from the fortress and canabae stand out for their relative quantities of Ebor ware vessels.Significantly, the North African-style casserole and lid ensemble, 96 Monaghan 1997, 1002, 1150.flagons (FE and FT).Whilst the u-profile bowls were produced in quantity, as evidenced by assemblages at 21-33 Aldwark, their thin distribution outside production contexts precludes further observations on their consumption.
To probe the connections of late Ebor types highlighted in CA in more detail, TABLE 5 details the contextual associations with selected pottery vessels and finds from published assemblages. 97This is possible by cross-referencing the contexts of published vessels and finds, enabled by the exemplary recording of contextual information in the fascicules of the York Archaeological Trust.Unfortunately, many small-finds data from excavations in York remain unpublished at the time of writing, especially for a range of important sites in the colonia, meaning that comparisons are more limited than desired.Some of the material considered, especially from the fortress, derives from redeposited dumps of Severan material in later contexts, which is not unusual at long-lived urban sites like York.For this reason, obviously intrusive material is excluded.
The contents of TABLE 5 mainly concern what might be thought of as assemblages produced by the primary consumers of the late Ebor repertoire: the military occupants of the fortress, also likely responsible for the configurations of redeposited fortress refuse making up the artificial terrace at Bishophill Senior.The list of associated finds includes lamps, recreational objects, considerable quantities of glassware and even a counterfeit denarius of Severus.The associations of pottery types and finds feature remarkably consistent elements.The direct association of North African-style casseroles and lids occurs in three assemblages: one each from the sites of 9 Blake Street (3268), York Minster (PH703) and 37 Bishophill Senior (10849).In two cases, the casseroles are recorded with exterior sooting, highlighting their likely use for cooking, irrespective of the absence of ceramic braziers.The most 'North African' assemblage is that of Blake Street 3268, which includes all three North African-style vessel types, a head-pot fragment of Caracalla and pieces of lorica squamata scale armour.Another assemblage from Blake Street includes North African-style casseroles and dishes (3223).Late Ebor vessels with possible northern Gallic links are missing from these assemblages, but are found in association in a couple of assemblages lacking North African-style forms at Bishophill (10157, 10171).While this apparent avoidance of northern Gallic and North African-style types is not universal, it strengthens the suggestion that the late Ebor repertoire was composed of distinct suites with different functional or culinary uses.
Of the associated pottery forms in TABLE 5, Ebor ware flanged bowls (BF4; FIG.7) appear in two of the assemblages with sooted casseroles from 9 Blake Street, reinforcing the suggestion of a culinary connection with the suite of North African-style vessels.Colour-coated beakers and samian cups and bowls appear in most assemblages.While the fortress evidently received proportionally less samian and imports than parts of the colonia in the early third century, the recurrent associations of late Ebor types with imported and fineware beakers and cups hints at a connection with communal alcohol consumption.This link is reinforced by the prevalence of glass beakers or cups in each of the selected fortress assemblages from Blake Street and York Minster in TABLE 5. Of other recurrent finds, gaming counters made of various materials, including glass, bone, samian and Ebor ware, appear in four assemblages.On the basis of the various repeated finds configurations, an image of the detritus of military personnel eating, drinking, gambling and relaxing together is inescapablewhether close to barracks, the military baths or perhaps even as working parties involved in the construction of the artificial terrace at Bishophill.A casserole set over heat seems well suited as a focal point for groups of soldiers socialising and sharing food.This kind of practice might have served to foster regimental esprit de corps by encouraging eating together as equals, especially at a time when the garrison at York had undergone considerable upheaval and change.Moreover, the lack of 97 Finds data are drawn from MacGregor 1978; Cool et al. 1995; Carver 1995 evidence for the civilian emulation of the full suite of North African-style vessels further underlines the possibility that it was used mainly by rank-and-file soldiers, and was therefore less suitable for imitation in the affluent civilian areas south-west of the Ouse. 98nother important feature of the contextual associations of late Ebor vessels with North African genealogy in York is their apparent close replication of associations of vessels and finds in assemblages in Africa Proconsularis.The cemetery of Pupput, located 70 km south-east of Carthage, is highlighted above for example vessels of strikingly similar design to those in late Ebor ware (FIGS 4 and 5).In this cemetery, equivalent types to those in late Ebor ware constitute the most common vessels selected for deposition in graves from the late second to early third century, namely dishes (Hayes 181), casseroles (Hayes 183, 184, 197 and Bonifay culinaire 25/Jdidi3) and lids (Hayes 185, 196), in addition to lamps and occasional vessels in African red-slip ware. 99These configurations strengthen the possibility that not only were specific pottery styles transplanted from North Africa, but so too were combinations of vessels characteristic of practices rooted in North African traditions.It is noteworthy that the ceramic braziers hypothesised by Swan to have been crucial to North African culinary practice are absent from the published graves at Pupput.If the funerary selection of vessels at Pupput was governed by idealised notions of dining and commensal hospitality as commonly seen elsewhere in the Roman world, the brazier was seemingly not considered to be an essential part of such idealised repertoires.While this parallel is suggestive, it should be noted that the most frequently occurring pottery type in the Pupput graves is the least common of the types produced and consumed in York: dish Hayes 181/Monaghan PA2.Likewise, equivalent combinations of vessels have yet to be found in a funerary context in York, which might be expected if the city was home to a diasporic population from Africa Proconsularis.
What do the mass of contextual associations of vessels and other finds reveal about the uses and significance of late Ebor ware in York?While the available evidence privileges military contexts, this is far from direct evidence of pots being made by Africans for Africans.Although the presence of North African soldier-potters offers the simplest explanation for the production of the repertoire in the canabae, there is no way of knowing if the consumers were North African (whether by birth, ancestry or self-perception).However, several factors point to the North African-style elements of the late Ebor repertoire being used as a distinct suite of 'barracks ware': the collective production and simultaneous emphasis on multiple North African-style types in waste dumps at Peasholme Green; the general military distribution of late Ebor ware in York; the repeated contextual associations of functionally linked types in military locations; and the consistent qualities of associated finds, most notably linked to socialising practices of drinking and recreation.This interpretation is supported by finds of equivalent vessels in other military contexts in northern Britain, 100 but does not presuppose the material was used only by soldiers, and especially not only soldiers from North African backgrounds.The recurrent production and consumption of North African-style types and combinations does, however, beg the question of whether this suite of vessels was understood as having North African or Mediterranean connotations.This question may be explored fruitfully through the application of organic residue analysis, 101 to determine whether the pots were used to prepare cuisine that was alien in York or, indeed, the same kind of food made in regular Romano-British pottery vessels.Indeed, the absence of ceramic braziers from York points towards cooking using casseroles on the hearth, perhaps 98 By contrast, earlier shifts in conspicuous pottery use by local elites in the north-western provinces were often inspired by fineware types favoured by Roman officers (Pitts 2019, 165-205).99 Ben Abed and Griesheimer 2004, graves 672, 135, 137, 1165, 134, 138, 196, 605, 1166, 647, 1169, 637. 100 For example, Swan 2008, 63, on finds from Carlisle.Much of the pottery with supposed North African influences on the Antonine Wall is stylistically different from late Ebor ware, for example Keppie 1985, 76-8; Swan 1999.  10Cramp et al. 2011.involving thick, dry stews, as suggested by experimental work. 102Either way, it remains highly likely that the North African-style ensemble was introduced to York as one of many changes brought about at the time of the Severan campaign, being popularised among the military community in this brief historical moment as a distinct cooking suite used by soldiers brought together from a variety of cultural backgrounds.

CONCLUSIONS
By the early third century A.D., Roman York was at the height of its importance as an imperial centre, hosting the emperor and his entourage in A.D. 208-11 and witnessing the culmination of changes to a rapidly developing cityscape begun at the end of the previous century.Urban history, of course, is much more than a story of powerful individuals and changing topography.This article has drawn attention to far-reaching changes that took place in Roman York's objectscape, focusing on an eclectic but short-lived suite of locally made coarseware vessels that took inspiration from distinct objectscapes elsewhere in the Mediterranean and north-western provinces, and its use in conjunction with a larger constellation of artefacts with a variety of local, regional and interprovincial origins.
From an objectscape perspective, this study reasserts Swan's hypothesis of a direct link with contemporary Tunisian production and consumption at the start of the third century A.D, especially for the novel types of casserole and probably also lids and dishes, likely introduced to York for the first time by soldier-potters from North Africa.Combinations of equivalent vessel types are common in contemporary funerary assemblages in Africa Proconsularis, but so far have only been found in military contexts in York.Whilst it is probable that other types in the late Ebor repertoire were influenced by northern Gallic and Rhenish objectscapes, this evidence is of lesser consequence in view of its smaller quantities and thin distribution in York.Suites of North African-style vessels dominate key production waste dumps from the likely military-controlled canabae, and the same configurations of vessels are repeated within consumption waste from the fortress.Outside the fortress, North African-style types are inherently diffuse and seldom thrown away together except in contexts that have been convincingly interpreted as redeposited fortress waste.Further contextual associations of the North African-style late Ebor suite of vessels reveal consistent and repeated connections with finds and pottery of a related functional character, most notably comprising other eating wares, sometimes with Mediterranean genealogy, high-quality glass and fineware drinking vessels, oil lamps and gaming counters.Taken together, the suite of objects is suggestive of the debris of communal eating, drinking and recreation by members of the military community.Rather than denoting the exclusive presence of African soldiers, a working hypothesis is that the North African-style suite presented an innovative culinary technology for soldiers brought together from a variety of cultural and geographical backgrounds in the unique historical circumstances of the Severan military campaign in northern Britain.
Lastly, this study permits some observations on the use of material culture, especially pottery, to shed light on issues of mobility and cultural diversity in the Roman world.This topic has been fraught with difficulty, in large part because of the implicit assumption that objects can be interpreted as markers of group identity and ethnicity.As the evidence of 'African-style' pottery demonstrates, material culture works in complex ways and is not well disposed to serve as simple proxy evidence for human mobility. 103Rather than ask what material culture can reveal about the movement of people, it is often more informative to ask questions of why new objects appear, what their impact is and why object styles and configurations (objectscapes) from different connected regions change in relation to one another.Acknowledging this complexity of material culture as a starting point can be tremendously valuable for understanding the rich milieu of Roman cultural dynamics and ought to provide important context to inform more holistic archaeological studies of human mobility.

FIG. 8 .
FIG. 8. Correspondence analysis: fabric supply in late second-and early third-century York, quantified by EVE and excluding samian ware; material from 37 Bishophill Senior included as supplementary data.
FIG. 10a.Correspondence analysis: the disposition of selected pottery assemblages determined by the comparison of vessel form; for full interpretation, compare FIG.10b.
Nevertheless, Monaghan argued that the closest parallels to the late Ebor forms are in fact Severan examples from North Africa, citing the presence of distinct concentric rings on the base interiors of some York vessels, also illustrated in examples from North Africa by Hayes. 58Comparing the late Ebor ware dishes to Mediterranean examples illustrated by Hayes and Bonifay (FIG.5, Monaghan form PA), it is clear that strong parallels exist with North African vessels, comprising fineware form Hayes 27/Bonifay sigillée type 13 and coarseware form Hayes 181/Bonifay culinaire B type 5. Again, the closest antecedents illustrated in FIG.
.66In contrast, butt-shaped jars of type JB1, with blockier rims, seem less closely tied to northern Gallic pottery traditions.The northern Gallic link for the u-profile bowls is also unconvincing.The closest parallels for the York vessels come from contemporary assemblages from Colchester, which may have developed from Rhine-Moselle prototypes. 67f u-profile bowl production in York must be explained by the movement of potters, individuals from Colchester seem just as likely as those from further afield.The last vessel types to be considered in the late Ebor repertoire are flagons with everted rims (FE) and trefoil mouths or pinched necks (FT) (see FIG.
72Pitts 2019.73 To investigate further, the rest of this study assesses wider patterns in the consumption of late Ebor ware.LATE EBOR WARE IN THE CONTEXT OF YORK'S POTTERY SUPPLYTo get a balanced picture of the distribution of late Ebor ware in Severan York, it is necessary to compare its presence at different urban locations with quantities of other pottery.On this subject, FIG. 7. Selected vessels related to the late Ebor repertoire (illustrations, types and find numbers afterWenham 1968;  Perrin 1990; Monaghan 1997).75Monaghan1997,874.the accounts of pottery supply in Monaghan's 1997 corpus are indispensable, including site-specific discussions of revised chronologies for important assemblages in older reports.Monaghan's key ceramic phase for the Severan period is 3a, c.A.D. 200-25.For the purposes of the present study, groups dating to this period are privileged alongside assemblages characterised by broader date ranges, where necessary taking in ceramic phases that cover earlier and later periods.76Summarydata concerning the most common fabrics from relevant pottery groups quantified by EVE (estimated vessel equivalent) are presented in TABLE

TABLE 2 .
PERCENTAGES OF MAJOR FABRICS BY SITE, DIVIDED BY CERAMIC PHASES FROM MONAGHAN (1997), QUANTIFIED BY EVE (material from 37 Bishophill Senior is quantified by min.no. of vessels)

TABLE 4 ,
FIG.10 reveals the results of CA conducted on the contextual associations of late Ebor vessels and other pottery forms.To ease

TABLE 4 .
THE DISTRIBUTION OF LATE EBOR WARE AND RELATED VESSEL FORMS AT SELECTED SITES IN YORK (quantification by numbers of illustrated examples and vessels listed in stratified groups)

TABLE 3 .
THE MOST COMMON AMPHORAE FABRICS FROM ROMAN YORK (all periods, quantified by % sherd count and % weight, with total number of sherds and weight (g) per site assemblage) 93These skeletons were not included in the sample from Roman York subjected to craniometric and stable isotope analyses inLeach et al. 2009.94 .