THE EVIDENCE FOR THE PRODUCTION OF GLASS IN

1 Jennifer Priee, H.E.M. Cool,

Romano-British Glass Project, Dept. of External Studies, Leeds University

!

Introduction Raw materials and glass frit

Discoveries from Wilderspool (May 1904, p. The basic raw materials for the manufacture 37-53), Caistor by Norwich (Atkinson 1932, p. 109) of Roman glass are sand, sodium and lime. and elsewhere have long indicated tha t a glass Deposits of these on a site should only be interpre­ industry existed in Roman Britain, and much new ted as the raw materials for glass production if evidence has come to light in recent years. This they are found in an undoubted relationship with has included furnace sites and associated finds, other evidence for it. No such deposits were found and assemblages of material in rubbish deposits on either of the sites where there is in situ eviden­ which, alth ough not associated with a furnace, ce of glass working (Mancetter and ). At appear to have been derived from a glass house. (Viriconium Cornoviorum) a small excava­ This paper gives a general overview of all this evi­ tion at the side of a road to the south-west of the dence with the exception of the new discoveries in town revealed an area of hearths and the type of London which are d ealt with elsewhere in this glass waste associated with vessel blowing. This volume (Shepherd and Heyworth). Much of this waste appears to be in situ at the edge of a glass material cames from excavations which are cur­ working site. White sand was also found, though rently being prepared for publication, and full subsequent chemical analysis has indicated that information will appear in those reports. Brief this was extremely unlikely to have been raw details of these sites are given in Appendix A and material for the glass produced (Sanderson et al. places mentioned in the text are located on Fig. 1. 1984, p. 60). The sources of evidence for glass production In antiquity the initial stage of glass making may be summarised as the ancient literature, epi­ was to frit the sand and the alkali. This involved graphy and archaeological remains. Only the last heating them together at a low temperature for a of these is relevant for the study of the industry in long period, before the resulting frit was ground Roman Britain, as no literary accounts or inscrip­ up and melted to form glass. A lump of frit is tions (except as mentioned below) survive. The more likely to be recognised as evidence of glass base patterns containing the letters CCV which are production if found on an archaeological site than, found on some square and rectangular botties for example, deposits of sand, but such finds are (Priee 1978, p. 70) are sometimes thought to repre­ rare. The only possible example of Roman date sent the initial letters of colonia Claudia Victricensis appears to be that found at Coppergate, York (modern Colchester) and thus to indicate manufac­ (Bayley 1987, p. 249 & 254). ture there, but this interpretation is questionable. The types of archaeological evidence likely There is no reason w hy each glass house to be available from glass house sites are well must have made its glass from the basic raw mate­ illustrated by late medieval illustrations such as rials; some may have worked from lumps of glass the Mandeville miniature (BM Additional Manus­ produced elsewhere. A triangular piece of glass 20 cri pt 24189, Folio 6, illustrated in Vose 1980, mm thick found at Culver St., Colchester may be Fig. 5). They may be divided into five categories. evidence for such a practice (Fig. 2.1). It has one These are deposits of raw material and glass frit, smooth and glossy surface and one pitted one and furnaces and annealing avens, crucibles and the may have been produced by pouring molten glass glass in them, tools, and glass waste from the pro­ into a tray. It is tempting to interprete this as part duction process. Much of the evidence from of a glass ingot but as no similar lumps have been Britain consists of isolated finds, so each category found in Roman Britain, this interpretation must will be reviewed separately, before looking at the remain tentative. complete assemblages from sites where material It is likely that much of the glass worked in was found in situ associated with a furnace. Roman Britain was made from re-cycled vesse!

23 , fragments. Literary sources indicate that cullet shawn by the dribble of glass adhering to the final was deliberately collected during the Roman per­ lining and the wasters, moiles and other glass iod (Leon 1941), and such a practice would explain waste found in its vicinity (see below). Cullet why so few restorable broken vessels are found in found with it suggested it had been in use during most military and civilian context§ in Britain and the mid second century, though from its stratigra­ elswhere in the western provinces, (Priee 1987, p. phical position it may have been of mid second 201). The large deposits of glass found in pits at century date or later. the canabae legionis at Nijmegen have been i.nter­ The association between pottery kilns and preted as cullet for use in glass dtaking (Isings glass furnaces is also recorded at Castor, Water 1980, p. 281). The large numbers of vesse! frag­ Newton where in the eighteenth century E.T. Artis ments found in the glass working assemblage at recorded a furnace with a crucible containing glass Mancetter suggests that cullet was the raw mate­ (1828, Pl. XXV, 4-5). It is noteworthy that glass rial fo r the industry there as well, and a similar waste found in a pit at Sheepen, Colchester was phenomenon was also observed at Leicester, also in the middle of an area of kilns (Allen 1983: Wroxeter and Sheepen, Colchester. Another col­ Appendix 1). lection of cullet appears to have been found at the A furnace similar in shape and size to the General Accident site at York where a group of one from Mancetter has also been found at vesse! fragments of fairly uniform size had been Leicester which was the main town of the civitas partly melted in controlled conditions. It may be, Corieltauvi (Coritani). It was found in one of the therefore, that basic raw materials were seldom shops associated with the forum in the centre of present on glass working sites. the town. A quantity of glass waste found with it suggested that it may have been used in the pro­ Furnaces and annealing ovens duction of glass vessels (see below), silver cupella­ tian may also have been carried out (Wacher 1974, The furnace is the most likely part of a glass p. 353). working site to be recognised archaeologically, These furnaces would have required cru­ though it will only have formed a relatively small cibles in which to melt the glass. The furnace part of such a site as the glass vesse! forming pro­ found at the Roman town of Caistor by Norwich, cess took place outside it. Sorne early representa­ by contrast, appears to have been a tank furnace. It tions of glass furnaces show the annealing aven was rectangular and measured at least 1.33 by forming the upper part of the furnace (A travers le 0.63m. Its lower part had been the furnace with a verre 1989, PL II 445, 50, 51), but this would not be secondary floor above it. Above this floor a band recovered from excavation where only the ground of fused glass c. 25 mm wide adhered to the side plan could be expected. of the structure (Atkinson 1932, p. 109-110, Pl. Apart from the recent discoveries in VA). On stratigraphical grounds it post-dated the London, six furnaces associated with glass pro­ early fourth century. duction have been recorded. Of these the earliest Glass furnaces have also been recorded at record occurs in the seventeenth century when the Wilderspool (May 1904, p. 37-58), but the precise antiquary John Conyers indicated that there was a status of these is open to question as none are furnace in London near the Fleet Ditch (Burnby recorded as being directly associated with melted 1984, p. 68). glass. The assemblage of glass from the site, howe­ The furnace (fig. 3, p. 30) at Mancetter is the ver, does indicate that sorne form of glass working best preserved example in Roman Britain and has was being carried out. Fragments from a glass fur­ the most complex history. Mancetter was the site nace have also been noted at Silchester but there of one of the main industries producing mortaria are no records of where in the town they were in Roman Britain (H artley 1973, p. 42), and the found, or what type of furnace they came from glass furnace was found during excavations of the (Boon 1974, p. 280). pottery kilns. The furnace is approximately circu­ lar with a flue to one side. It was made of clay and had been re-lined three times. In its final phase it Cru cibles had a floor of tiles and was oval internally. One of the most noteworthy features about this structure The size of the furnaces at Mancetter and was tha t it was very small. In Phase I it had an Leicester indicates that they would have only held internai diameter of 0.8 metres but by Phase 3 this one crucible. Too few examples have been found had been reduced to 0.51 by 0.34 metres. There to generalise on the size and shape of these. was no obvious structure that could be interpreted as an annealing aven in the vicinity, and the small The crucible recorded in the furnace at size of the furnace itself would appear to preclude Castor was flowerpot-shaped, and a black of glass the possibility that the oven was above it. That the found at Castor from the base of this or a similar furnace had been used for glass production was crucible has retained an approximate base diame-

24 ter of 150 mm. A siinilar fragmentary crucible has end, closest to the paraison, as the compression also come from Silchester (Boon 1974, p. 280). may have occured when the blowing iron was put These appear to have been purpose-made contai­ down. lt is probable that these came from vessels ners, although domestic pottery was sometimes with hot finished rims as these would have been used. Fragments from a third century flanged, oxi- removed much doser to the blowing iron. dised bowl containing a thick layer' 1of blue/ green Cylii1drical moiles have been found in large num­ glass have been recovered from Deansway, bers in the assemblages from Mancetter, Leicester Worcester. The crucible fragments n~covered from and Wroxeter, and also in the assemblage of glass Coppergate, York may also have been of domestic waste from Sheepen, Colchester. pottery (Bayl ey 1987, 249; Figs 2 & 3). Lid-like moiles are cylindrical in their upper parts and expand out to wide lower bodies wh:ich Tools have cracked off edges (Fig. 2.4 - see also Priee 1975, Fig. 4 for examples from Merida). They come Tools and other equipment attributable to from vessels on which the rim was cracked off glass working have not been found on Romano­ after annealing, and represent the portion of the British sites with the possible exception of flat paraison between the blowing iron and the final slabs of stone at Wilderspool that may have been rim. A vesse! from Trier, published as a flask, used as marver blacks (May 1904, p. 50). A sim:ilar appears to be an annealed paraison for a hemis­ scarcity also exists in the rest of the Roman world. pherical eup which still has the lid-like moile atta­ To a certain extent this is not surprising. By anale­ ched (Goethert-Polaschek 1977, p. 153 no. 917, Taf. gy to modern practice, it is likely that some tools 59). A second moile of this variety from Trier was were made of wood and these would not normally found as part of the grave goods in an inhumation survive. Others such as iron shears and pincers belonging to the first half of the fourth century may survive but can have other uses, and when (Goethert-Polaschek 1977, p. 252 no. 1485, Taf. found in isolation are not necessarily indicative of 16.176d). The only lid-like moile to have been glass vessel production. The one tool unique to identified from Britain cornes from Culver St. this industry is the blowing iron but very few of Colchester, though a fragment in the Sheepen these have been recognised from the Roman per­ Colchester group may have come from another. iod. The only examples known are the blowing This scarcity may in part be due to the difficulty of irons found with pincers, moiles and cullet at recognising examples broken into small frag­ Merida, Spain (Priee 1974, p. 82, Fig. 5), and that ments. from Salona (Auth 1975, p. 167). The second category consists of three types of waste that appear to be the regular by-products Moiles, other glass waste and wasters of particular processes. One, which may be called a roundel, is circular or oval with a diameter of c. The commonest indications of glass wor­ 10 mm (Fig. 2.5). These have convex, shiny upper king are fragments of glass waste. Under normal surfaces and concave, irregular lower surfaces glass house conditions the majority of these would which often incorporate impurities. The process have been re-melted and used again, so the volu­ represented by these is not immediately apparent. me of those fotmd need not be directly related to Fragments with pincer marks and a clipped edge the scale of production on the site from which they (Fig. 2.7) are the waste from attaching thick trails are derived. It is possible to distinguish eight dif­ of glass and may be associated with the produc­ ferent types of glass waste which have been grou­ tion of handled vessels. The third type of waste ped into four categories. consists of a fragment with circular base with a The first category consists of the fragments diameter of c. 20 mm, which has been drawn out called moiles, which are the waste from around and twisted in its upper parts. The under surface the end of the blowing iron. Two different varie­ of the base is concave and sometimes has iinpuri­ ties of moiles have been distinguished. Cylindrical ties (Fig. 2.8). It is possible that this twisted waste moiles are commonest. Most have a diameter of cornes from the production of handles or of twis­ 20-30 mm, are full of bubbles and/ or iinpurities ted rads, though it may also have been removed and sometimes have ridges in relief on the inter­ from other products. This is the rarest of these ior. Two different types of 'rims' occur regularly; three types of waste. It has only been found in one one has a sharp edge often with the appearance of of the waste assemblages discussed below a concave bevel towards the interior (Fig. 2.2), the (Wroxeter), though isolated finds have been found other appears to have been compressed (Fig. 2.3). at Culver St. Colchester and Churchgate, Leicester These moiles are the waste glass left on the blo­ ( Museum no. 116/ 1962). wing iron after the paraison had been detached. It The third category consists of three types of seems likely that fragments with a sharp edge are waste that may be viewed as more accidentai by­ from the end around the blowing iron, and those products. They consist of the trails and thin rods with the compressed edge are from the opposite that form when molten glasS" drops off the gather;

25 the rounded (Fig. 2.6) and fractured lumps created assemblages. That for Wroxeter has also been by plunging blowing irons or the contents of cru­ included in the comparison because though no cibles into water; and the rniscellaneous melted furnace was found, it seems likely that the mate­ and heat affected lumps that cannat be more close- rial could have been in situ. ! y defined. r· : At none of the sites were there either raw The fourth category of waste is vessels that materials or frit, and at all three the industry becarne distorted during manufacture and were appeared to be working with cullet. Nor were any discarded (wasters). The assemblagr at Mancetter glass working tools or crucibles found, although at included such rnisshapen and very bubbly vesse! Mancetter and Leicester it is clear that crucibles fragments, but none of the other sites dealt with in must have been used . This absence presumably this paper have produced such wasters. indicates the removal of the crucibles and tools These four categories of waste have diffe­ when the furnace was abandoned. rent levels of importance for interpreting an Only at Mancetter were wasters present assemblage in which they occur. The rnoiles defi­ which indicated the range of vessels being produ­ nitely indicate the occurrence of glass blowing. ced. These included rim fragments from collared The second category indicates that glass working jars (Isings 1957, Form 67b/c) and srnall jars with was taking place. The types of waste in the third funnel mouths, and bases from base rings perhaps category, however, could be produced accidental­ from jugs or bowls (Fig. 2.9-13). Very bubbly ly, and the same is true of 'wasters' as vessel frag­ blue/green body fragments also suggested that ments can be distorted by heat accidentally and sorne of the vessels were decorated with trails. The not just during manufacture. If these criteria are cylindrical moiles indicate the colours of the ves­ applied to the assemblages recovered, it becomes sels being produced on the three sites. In each case apparent that only at Mancetter, Leicester, most of the production was in blue/ green glass. Wroxeter and Colchester (Culver St. and Sheepen) At Mancetter there was also sorne production in is there any evidence for the blowing of glass ves­ colourless and in yellowish glass, and at Leicester sels. No rnoiles have survived in the assemblage colourless vessels were made. A t Wroxeter there from Wilderspool or that associated with the tank are sorne indications of production in light green furnace at Caistor by Norwich. The glass in those and yellowish glass, but the evidence for this is assemblages included melted lumps but not the inconclusive. more diagnostic fragments belonging to the The different types of glass waste (other second category of waste. Glass rnelting was clear­ than wasters) recovered from these sites are surn­ ly carried out at these sites but for what purpose is rnarised in Table 1. It is noteworthy that the total not known. The si tua ti on is corn pl ica ted at amounts recovered, rneasured by weight, are not Wilderspool because the assemblages contain great. This presurnably reflects the fact that, as sorne very bubbly and distorted base fragments noted above, under most circumstances this waste which it is ternpting to interprete as wasters. They would have been re-cycled and not discarded to suggest that glass vesse! blowing may have been form part of the archaeological record. taking place in the vicinity, but on their own are not sufficient proof of this. The composition of the waste assemblage is n ot the same on the three sites. Those from The glass working sites at Mancetter, Mancetter and Wroxeter are similar in that the Leicester and Wroxeter moiles make up approximately one-third of each and exarnples of waste from the second category Of the sites mentioned so far, the industry are relatively uncommon. At Leicester this second can only be studied in situ at Mancetter and category is much more strongly represented, with Leicester, an d it is instructive to compare their approxirnately equal amounts of roundels and

1 2 3 4 5 6 7 Total Weight

Mancetter 36% 5% 1% 0% 8% 18% 32% c. 1.2 kg

Leicester 15% 18% 18% 0% 4% 13% 32% c. 0.5 kg

Wroxeter 37% 2% 1% 5% 17% 19% 19% c. 1.0 kg

Table 1: Types of glass waste present at Mancetter, Leicester and Wroxeter 1 : Cylindrical moiles - 2 : Roundels - 3 : Pinched and clipped fragments - 4 : Twisted fragments - 5 : Lumps with rounded surfaces - 6 : Rods - 7 : Miscellaneous Lumps

26 clipped fragments perhaps reflecting that they stances, though the surviving evidence is frag­ were ba th associated with the same production mented, it may be a reflection of a thriving glass process. industry in Roman Britain. The reasons for the differences in the assem­ The locations of the furnace sites are very blage compositions are not clear. It l(tay be that the interesting. It seems reasonable to expect evidence different proportions reflect the pro~uc tion of dif­ of glass prod uction at towns Ii ke Colchester, ferent types of vessels. The possiblity that they Wroxeter and Leicester where there will have been indicate chronological differences jn workshop a sizeable market for the finished goods. Glass tradition can also not be ruled out. The Wroxeter working at small sites su ch as Mancetter and assemblage is of late firsf or second century date Wilderspool is superficially more puzzling, but and the Mancetter assemblage may be dated to the there are grounds for thinking that influences mid seco"nd century. The glass waste fro m o ther than the presence of a market may have Sheepen, Colchester, which hasan overall compo­ affected the choice of these locations. sition similar to that from Mancetter, is thought to In thr ee cases (Mancetter , Castor and date to the mid second century or slightly later Sheepen Colchester) there is a certain association (Allen 1983, p. 772). The Leicester assemblage, by of glass furnaces with pottery kilns. Such an asso­ contrast, is later and probably of third century ciation would have much to commend it. Glass date. More dated glass waste assemblages are nee­ production needs good supplies of raw materials ded before interpretations can be made. and fuel, as well as a good access to a market. Pottery is also a high temperature industry and Conclusions wherever kilns are located there is likely to be a good supply of fuel. The pottery industry needs Although the evidence for the glass indus­ good access to its markets and a glass house loca­ try in Roman Britain is scattered and fragmentary, ted in the vicinity could make use of the marke­ it is possible to d raw sorne general conclusions ting system used for the pottery, whether or not from it. It is clear that glass vessels were being this is envisaged as the regular consignment of produced in Roman Britain, though it is difficult large loads to traders, or visits from itinerant ped­ to judge how widespread such production was lars replenishing their stock, or some other mecha­ because by its very nature much of the evidence nism. lt has already been noted that, the glass has not reached the archaeological record. One of houses in Roman Britain were probably making the most diagnostic classes of evidence is glass use of cullet as one of their main supplies of raw waste, but under most circumstances this will be materials, and it is possible that the same traders re-cycled. This is very unlike the situation in the or pedlars who sold the vessels may also have pottery industry, for example, were wasters from been involved in the collection of cullet. This the kiln have to be discarded as they are of no use. would thus have been returned to the glass houses To continue the comparison between the pottery through the agencies that were also distributing and the glass industries, it is also much more diffi­ the vessels. cult to locate Roman glass house sites than it is to If this madel is correct there may have been locate kiln sites. Not only will there be an absence many small scale glass production sites associated of the equivalent of a dump of wasters, but the with pottery incl us tries. It is important th at furnace itself will be much smaller than a kiln. It is archaeologists excavating kiln sites are aware of probably not without significance that ail the fur­ this possibility, as the recognition of glass produc­ naces that have been located have been accidentai tion waste may alert them to the presence of fur­ discoveries, in the sense that in none of the cases ther glass houses in the future. was the site was being excavated in the expecta­ tion of finding a glass house. In these circum-

\ 27 \ AppendixA: Acknowledgements Unpublished sites mentioned in the text Much of the work for this paper has been The excavation reports of following sites are undertaken by the Romano-British Glass Project, currently being prepared for publiç-ation, and full funded by Historie Buildings and Monuments details of the glass assemblages will be published Commission (). Access to the glass from in those. Wilderspool and Caistor by Norwich was arran­ i ged by Mr. A. Leigh of the Warrington Museum -Colchester, Culver St. Excavations in 1981 - 1985 and Art Gallery and Miss B. Green of the Norwich by the Colchester Archaeôlogical Trust. Castle Museum respectively. Mrs. K. Hartley, Dr. - Leicester, Blue Boar Lane. Excavations by Mr. J. J. Haughton and the Archaeology Section of the Wacher in 1958 (Journal of Roman Studies 59 HWCC made available the material from their (1959), p. 114) currently being prepared for publi­ excavations in advance of publication. Mr. N. cation by Mr. N. Cooper of the Dept. of Archae­ Cooper gave us much help with the glass from ology . Leicester, and Dr. D. Allen discussed with us the - Mancetter, . (Sites 7 and 7 / 20) assemblage from Sheepen, Colchester. We are Excavations by Mrs. K. Hartley in 1964-5 and greatly indebted to ali these individuals and orga­ 1969-71. A brief note about the furnace is given in nisations for their co-operation. HEMC also V ose 1980, p. 132-3. wishes to thank Mrs. F. Pate for help with the - Worcester, Deansway. Excavations in 1989 by translation of the paper delivered in Rouen. the Archaeological Section of the Hereford and W oreeste r County Council. - Wroxeter. Excavations in 1972 by Dr. J. Haughton. Brief notes in Britannia, IV, 1973, p. 287 and Archaeological News Sheet. 16, 1973, p. 17

Abstract Résumé

The evidence for glass working and the pro­ Les témoins du travail du verre et de la pro­ duction of glass vessels in Britain during the duction de vaisselle de verre en Angleterre durant Roman period is reviewed. Pive categories of la période romaine sont passés en revue. Cinq archaeological evidence from glass houses are types de vestiges archéologiques provenant d'ate­ identified - raw materials, furnaces, crucibles, liers de verriers sont identifiés : matières pre­ tools and glass waste. The last category is divided mières, fours, creusets, outils et déchets de verre. into eight different types, sorne of which are dia­ La dernière catégorie est divisée en huit types dif­ gnostic of glass blowing. It is suggested that there férents dont certains sont en relation directe avec may have been many small scale glass production la technique du soufflage. si tes associa ted w i th other high tempera ture Nous suggérons qu'il a pu y avoir beaucoup industries such as pottery, and often using cullet de sites de production du verre de petite dimen­ as the raw material. sion, qui utilisaient souvent du groisil comme matière première et qui étaient associés à d'autres arts du feu comme la poterie.

28 K EV Furnace *& Cruclble ••11 • B low lng waste

••10 ••8

Fig. 1 : Location map of sites mentioncd in the text (1 - Caistor by Norwich. 2 - Castor, Water Newton. 3 - Colchester. 4- Leicester. 5- London. 6- Mancetter. 7- Silchester. 8- Worcester. 9- Wilderspool. 10- Wroxeter. 11 -York).

29 ~ ~ 1 1 1 'V -v- 3 {.______) 1 4

_.L:J -0- 5 1 1 b 1 1 /fl - ~ - ij 9 ~ ~~-, - ~ - ·.::: ... ~ti.:· ,. :r · - ... 6 ~ 7

~1 2

~ 11

E""""'3 E"'3 1 1 ems

Fig. 2 : Examples of glass waste.

Fig. 3 : The glass furnace at Mancetter (scale in imperial measurements).

30