Buildings and services (Finds Functions 9 and 11)


The fabric of the buildings on the site was represented by a variety of materials; burnt daub, tile, mortar and opus signinum, wall plaster, tesserae and window glass, as well as the fastenings and fittings used to hold them together. Some of this material existed in large quantities; over 7 tonnes of Roman tile and 316kg of burnt daub were collected. In contrast, there were very small assemblages of wall plaster and window glass.

The nature of the buildings of Essex has always been influenced by the available raw materials. In particular, the county is lacking in building stone, though clay is readily available. Timber, wattle and daub, and tile were the normal building materials in Roman Essex. All good building stone, and much of the coarse stone rubble, derived from outside the county, and consequently little is found outside the major towns. Elms Farm is no exception with virtually every building being of timber-framed construction (as evidenced by the post-holes and beam-slots.

Some of the nails recovered may have been used in building construction, but the majority were probably too small for heavy structural timbers. The walls of most buildings on the site were made from wattle and daub, and some of this appears to have been plastered. The use of stone for structures on the site was not common, and there is little evidence for the use of shaped building stone. Stone rubble was used in the foundations of the temple precinct wall, in the plinth of the temple, as footings for Building 63, and in some oven structures and hearths. Tile was also used as a structural element (apart from in roofs). It seems to have generally been used in the same sorts of structures as the building stone, that is in ovens and hearths, as packing in foundation trenches and post-holes, and in the precinct wall.

Some evidence for the use of glass in windows was recovered, consisting of ninety-six fragments, all from post-conquest contexts. Roman window glass in Essex, outside Colchester, is fairly uncommon. It is reasonable to suggest therefore, that a substantial building existed not too far from the excavated area at Elms Farm. A small amount of painted wall plaster was recovered. It came from an area adjacent to the temple precinct, and may have come from the temple or could conceivably have come from a building located outside the western edges of the excavated area. Little evidence of built interior floors survived. Some of the opus signinum may have been from floors, and there were sufficient loose tesserae to cover a total area of c. 4.5 x 2m.

Most of the roofs would have been thatched. The very large assemblage of Roman brick and tile (over 7.1 tonnes), included a substantial amount of roof tile but at least some of this had been used as a general building material in structures such as ovens. There were no collapsed roofs, or large dumps of roof tile to indicate which buildings may have had tiled roofs. A number of 'chimney pots' were present; however, in the absence of any evidence of sooting their function is by no means clear.

The artefactual evidence for water management on the site consists of fragments of ceramic pipes, and a number of iron water-pipe collars from wooden pipes. The water-pipe collars came from a late 3rd to 4th-century watercourse in Area R. The bottom of Ditch 25271 was lined with wooden timbers, and it was clearly not just a boundary ditch, but part of an organised water management system, draining into the watercourse to the south. Given that other artefactual evidence (the window glass and tile in particular) suggests the presence of a major building in the vicinity, it seems reasonable to suggest that these two features form part of a water management system associated with that building, possibly a bath-house, located in the unexcavated area immediately to the north.

Group of iron water-pipe junction collars

Not illustrated

A number of fragments of water-pipe junction collar came from contexts 12026, 12029 and 12036. All are fills of ditch terminal 12027 (Ditch 25270, Group 968, Late 3rd to 4th-century, Period 4-5). 12029 is a dump of pot in the top of the ditch. There are joins between contexts 12026 and 12036, the breaks of which are fresh, indicating some mixing of the finds from the two contexts, and it seems reasonable to suppose that this group of collars is a single deposit. In addition, 11000 SF4382 must be part of this group, even though it is ostensibly from a different area of the site. The size is the same, the soil matrix is similar, and the SF number is close to one from 12026. Presumably it was mis-numbered.

The fragments are c. 26mm wide and have a very low external central flange. Mineralised wood is present on one or both faces. On the inner surface, the butted ends of the pipe show clearly on some pieces. The diameter is difficult to measure accurately in all cases, as most pieces have some distortion from a true circle, but the range is between about 120mm and 140mm internal diameter. Assuming that these are all fragments from the same collars, the minimum number of collars present was calculated (using chord measurements) to be three.

The objects were not recognised as water-pipe collars during excavation, so not all the small finds in the group were plotted. The plotted fragments fall into three groups, whose positions suggest that the water pipe was in situ, and that the length of the wooden pipes was c. 1.8-2m.

Ceramic pipes

There were fragments of nine pipes or tubes from eight contexts. They are in a fairly fine tile fabric with sparse sand, except for one sherd (Context 10028), which is sandier, and the illustrated piece from 14985, which is in a pottery fabric, and has an octagonal section. It is unlikely that any of the pieces from different contexts are from the same tubes, as the internal treatment varies.

Apart from the pipe from 14985, the tubes were coil built, the junctions of the coils being visible on the inside of many of the pieces, and most sherds having broken along the line of the coil. The insides were not sanded, although in the case of the piece from 10104, the tube was probably placed vertically on a sanded surface to dry. The outsides were fairly well finished, with no trace of the coils visible. The octagonal pipe from 14985 appears to have been slab built.

The external diameters of the tubes were between 100mm and 156mm, with slight evidence for tapering on some pieces. The walls were 10-20mm thick. The octagonal pipe was somewhat smaller in diameter, but had a comparatively large bore since the walls were thinner than the tubes in tile fabric. Too little survives to be certain of the precise form of the tubular pipes. The octagonal pipe is paralleled by a hexagonal one from York (Brodribb 1987, 85), although the latter example is in a somewhat thicker tile fabric.

The contexts containing pipe fragments vary in date from Early to Late Roman, and are concentrated (if one can say that about such a small group) in Areas E and F, which produced six out of the nine fragments. The other three fragments were from Areas G, K and L.

Pipes are discussed by Brodribb (1987, 84) and Wilmott (1991, 52), who note various uses for them. They were most commonly used for conveying water, either as an in-pipe or a drain, although it is rare for them to be found in situ, or for there to be any other evidence of how they were used, and the Elms Farm examples are no exception. Indeed, it is possible that they were not specifically brought to the site for use as pipes, but were brought in as second-hand building material, as most of the tile appears to have been. However, we know from the evidence of the iron junction collars that wooden pipes were used on site, and it is certainly conceivable that ceramic pipes were used as well, presumably somewhere in Areas E and F, starting before about AD 160. The small amount of pipe retrieved suggests that they were not used extensively on the site.

Figure 482: Ceramic pipe, 1-2

1. Six joining pieces of a tube, forming 50% of the circumference. The flat end is sanded, presumably having been stood on a sanded surface to dry. The inside is fairly well smoothed, and is possibly partly slab built rather than coiled. An implement seems to have been used for the smoothing, as well as the fingers. External diam. 124mm, Th. 15-20mm, Ht 137mm. Wt 734g. Layer 10104, Group 813, Area F, Period 4

2. Fragment of a pipe with a hexagonal section, in a pottery fabric (sandy grey ware: Going Fabric 47: T.S. Martin, pers. comm.). The outer surface is well finished, the inner somewhat irregular. The width of the pipe was c. 72mm, with an inner diameter of c. 62mm. L. 105mm. Wt 86g SF8423, Fill 14985, Well 14984, Group 710, Area L, Period 4-5

Mortar and opus signinum by Ros Tyrrell

Cite this as: Tyrrell, R. 2015, Mortar and opus signinum, in M. Atkinson and S.J. Preston Heybridge: A Late Iron Age and Roman Settlement, Excavations at Elms Farm 1993-5, Internet Archaeology 40.


Just thirty contexts resulted in the recovery of 42.180kg of mortar. This is a small quantity for a site of the size of Elms Farm, but few of the buildings and other structures employed mortar in their construction. This is normal for a rural Roman site in Essex, where the lack of good-quality building stone meant that timber, wattle and daub were the usual construction materials.

Figure 483: Distribution of mortar (weights in kg)

Area J produced 63% of the material, all from contexts associated with the temple precinct (Figure 483). Samples were taken from the plinth of the temple (5950 and 5588) and from the precinct wall (22079), and 2kg came from the construction (5421) of Building 63, in the same area. The 0.7kg of material found in Period 4-6 pits (15811, 10953, 10995, 7114, and 14529) may have come from the temple. Tile-built ovens produced 62% of the mortar from Area M and 99% from Area K. These structures contained some mortar but do not seem to have been fully bonded.

Table 84: Quantification of mortar by period (weights in kg)
Period 3 3B-4 4 4-5 5 5-6 6 0
Weight 15.866 1.815 0.905 2.500 12.849 1.500 6.060 0.685

An unspecified quantity of mortar was recovered from the 1972 excavations (Wickenden 1986, 21), from a 2nd-century pit. This pebbly white lime mortar had evidently been used as roof flashing, since the material retained the impression of imbrices. It is interesting that there is material deposited some distance from the temple area, but the location of the building that the mortar came from remains unknown, and it may not have been any of the buildings in the area excavated in 1993-5.

Opus signinum

All the material was macroscopically examined (see report by Graham Morgan for the scientific analysis). Opus signinum is considered to be mortar with varying quantities of crushed brick and/or tile added during mixing. The resulting material was used structurally, for instance, in roofs, walls and mosaic bedding, and decoratively for floors and bath linings. Roman builders used such material in conditions where damp was a potential problem.

While the total weight of opus signinum recovered was 101.75kg, 60.79kg came from a single pit in Area I (13358). The weight from this pit is misleadingly large, as the opus signinum was firmly attached to fragments of bonding tile, which had to be weighed with the opus signinum. Despite this, there is still a substantial quantity of material from Area J, all from contexts associated with the temple precinct. With the exception of post-pads 18924 and 21978, and wall robbing 21997, few of the deposits were found in situ, and the mid- to late Roman dates of the pottery found with them suggests that much of the material may be redeposited.

Table 85: Quantification of opus signinum by area
Area I J K N R
Weight (kg) 61.010 40.340 0.210 0.100 0.090
Table 86: Quantification of opus signinum by period
Period 3 4 4-5 5 5-6 6
Weight (kg) 0.110 5.140 0.090 61.790 0.220 34.400

The Elms Farm opus signinum is a very coarse form of the material, having larger than usual inclusions of crushed brick (10-15mm, as against an average of 1-5mm). The mortar that forms the matrix looks very similar to the plain mortar from the site. Apart from the samples sent for analysis, none of the material shows any definite signs of having a surface such as would be found on floors or walls. It seems to have only been used in construction. However, large redeposited fragments from pit fill (13216) in Area I displayed two or three layers of possible floor surfacing, hinting at the past existence of a substantial Roman building somewhere near the site.

Only three features from Drury's 1972 excavations (Wickenden 1986, 21) yielded opus signinum, so little meaningful comparison can be made between the two sites.

The assemblage, though limited to a small number of contexts (sixteen in total) does indicate the presence of a relatively substantial and sophisticated building probably somewhere within the settlement. As an aspect of evidence for buildings, their use, status and construction, this small assemblage can be seen to have limited value.

Mortar and opus signinum analysis by Graham Morgan

Cite this as: Morgan, G. 2015 Mortar and opus signinum analysis, in M. Atkinson and S.J. Preston Heybridge: A Late Iron Age and Roman Settlement, Excavations at Elms Farm 1993-5, Internet Archaeology 40.

Samples were examined microscopically and chemically to determine and compare their compositions. They were all lime based with varying aggregates, some mainly sand and gravel, the rest mainly crushed orange/red brick or tile.

Table 87: Mortar descriptions
Context Description
21814, Group 403 Coarse cream sand and gravel mortar lump with tile traces.
21942, Group 439 Coarse cream sand and gravel, 40mm thick, pale brown limestone fragments.
18924, Group 456 Pink coarse sand and tile mortar.
21724, Group 439 Pink/red crushed tile floor? 22-25mm thick.
21797, Group 403 Pink/red crushed tile mortar with some gravel. Flat surface, floor?
21799, Group 440 Pink/red crushed tile floor? 25mm thick.

Chemical separation showed that the mortars were composed mainly of crushed brick or tile with varying amounts of flint quartz and quartzite, in that order, for the gravel-sized components and round to sub-angular quartz sand in sand-sized particles. Those samples containing mainly crushed tile as the aggregate also contained sand which in part was derived from the tile itself, being freed as the fine sandy tile was crushed. In the gravel and sand mortars flint was the major component in the coarser grades with sub-angular quartz being the bulk of the sand-sized particles.

Table 88: The components of the material
Context Gravel >2mm Sand 0.15-2mm Silt <0.15mm 'Lime' % Comments
21814, Group 403 40 56 4 21 Sand/gravel
21942, Group 439 42 54 4 30 Sand/gravel and limestone
18924, Group 456 51 39 10 29 Tile
21724, Group 439 75 17 8 22 Tile
21797, Group 4033 64 27 9 22 Tile/gravel
21779, Group 440 73 17 10 23 Tile
Figure 484: Particle-size distribution in tile-based mortars

The particle size distribution analyses are shown on Figures 484 and 485. It can be seen that the sands in the tile-based mortars and gravel-based mortars are significantly different, showing different sand sources, but that there is a certain amount of the gravel/sand-sized particles also in the tile-based mortar.

Figure 485: Particle-size distribution in gravel-based mortars


The 'lime' values, which approximate to the original lime content of the mortars, are about right for normal sand- and gravel-based mortars but somewhat low for opus signinum tile-based mortars, when compared with those analysed in a national survey of Romano-British mortars and plasters (Morgan 1992b). It may be that they are poor imitations of opus signinum, they are deliberately low in lime to make them harder for floor use or that crushed brick or tile was simply used in place of sand and gravel. The higher 'lime' content of 21942 is in part due to the presence of limestone fragments.

Painted wall plaster by Ros Tyrrell

Cite this as: Tyrrell, R. 2015, Painted wall plaster, in M. Atkinson and S.J. Preston Heybridge: A Late Iron Age and Roman Settlement, Excavations at Elms Farm 1993-5, Internet Archaeology 40.

Eight contexts produced 3.32kg of painted wall plaster, with a total surface area of 1432cm². In particular, it was found in a cleaning layer in Area K, 4881, three re-cuts of a pit (14097, 14098 and 14099) and pit 10910 in Area N. The majority of the material came from Area K, mostly from 4881. The relationship of this context and the other plaster from Area K was not recorded. Area N produced only 33cm². The nature of the designs does not allow the painting to be easily dated intrinsically, but is not inconsistent with the 3rd-century date of the associated pottery.

Table 89: Quantification of wall plaster
Context Feature Group Feature type Period Site Area Area (cm²)
4881 - - Cleaning layer Undated K 1122
14083 14097 4019 Pit 4-5 K 50
14043 14098 4019 Pit 4-5 K 4
14052 14098 4019 Pit 4-5 K 57
14089 14098 4019 Pit 4-5 K 17
14093 14099 4019 Pit 4-5 K 149
10891 10910 676 Pit 5 N 18
11139 10910 676 Pit 5 N 15

Despite the fact that the material from Area K is quite widely separated from Area N, it all looks similar enough to have come from the same source. Area J, which produced substantial quantities of mortar and opus signinum and includes the temple, has no painted plaster. Since the nearest Roman building beyond the site is apparently some distance away, it can only be assumed that it came from the immediate area of Elms Farm, and from a building redecorated or demolished by the early 3rd century. There is no evidence to suggest that the daub buildings were painted, although they could have been. Unfortunately there are no stone or wood impressions on the backing of the plaster to indicate the nature of the building that the plaster came from.

The Elms Farm plaster is well finished but there are no signs of polishing. The larger fragments have a maximum of three coats of plaster. The first is a coarse mixture with pieces of crushed brick and flinty debris. Next is a coat that has powdered, as well as crushed, brick added, giving it a more pinkish appearance. One of the fragments had broken at this layer and there are traces of an earlier phase of decoration consisting of a grey bead on a white ground. The finishing coat is a whitish sandy mixture with a few flecks of crushed brick. The inclusion of large quantities of crushed brick is usually an indication that there was a need for water/damp-proofing, as in a bath suite.

Colour decoration divides into two groups: plain colours; and colour combinations on a white ground, with bands, stripes and beads represented in various colours. Most of the colours are painted onto a white background with the exception of one fragment, which is pink with an orange undercoat. This method was probably intended to enhance the tone of the colour.

Although there is insufficient material to suggest any decorative schemes, the room was probably painted with large areas of plain colours divided into panels by contrasting stripes. With the possible exception of the beads, the design is purely geometric. The beads may have been placed at the corners of the panels, as in the reconstruction of the early 3rd-century wall decoration of Building XXII, Corridor 5 at Verulamium (Liversidge 1984, 126).

The painted plaster from this site is evidence for an unidentified Roman building of moderate status. It could have come from the temple, despite the fact that it was not found in the precinct.

No painted plaster was found in previous Heybridge excavations to compare with the material from Elms Farm. The geographically nearest comparable material is from the mansio at Chelmsford (Drury 1988, 86-90), which had similar panel type designs. These came from contexts dated c. AD 150 and were therefore in situ some time before that. In contrast to Elms Farm, where only eight contexts contained wall plaster, the temple at Ivy Chimneys, Witham, produced wall plaster from over a hundred contexts (Turner 1999, 198). The bulk of the Ivy Chimneys material comes from 4th to 5th century contexts and so is somewhat later than the Elms Farm material.

Plaster analysis by Graham Morgan

The samples were too small for full analysis.

Cleaning layer (4881) Area K

  1. Two layer plaster; grey or dirty white <0.05mm, on sandy white intonaco, 0.4mm, on coarse sandy white plaster with some tile pieces, 8mm on dirty white, <0.05mm on white intonaco, 0.1mm, on coarse pink sand and gravel 9mm thick. This is an example of a redecorated wall.
  2. White line on orange/red (brick dust), 0.1mm, on sandy white intonaco, 0.5-0.7mm, on sandy pink plaster, 4mm, on coarse sand, gravel and tile plaster, 20mm thick.

The daub by Ros Tyrrell

Cite this as: Tyrrell, R. 2015, The daub, in M. Atkinson and S.J. Preston Heybridge: A Late Iron Age and Roman Settlement, Excavations at Elms Farm 1993-5, Internet Archaeology 40.

The material described here is structural clay, accidentally burnt. Baked clay that had no apparent added inclusions and appeared to be burnt soil was not included in the assemblage. Pieces with no structural features are considered to be daub if other pieces in the same fabric had traces of wattling or similar structural evidence. The fragments derive from the use of hearths and ovens, and the burning of timber framed buildings that had wattle and daub infill (Drury 1978, 114).

A total of 13,035 fragments of burnt daub, weighing 329.664kg, was recovered from 659 features, with an average fragment size of 25g. Much of the material has been disturbed and redeposited, and is abraded as a result; however, wattle impressions and variations of surface treatment can be detected on some pieces. The largest assemblages were found in Area G, comprising 49% (by weight) of the total from the site (Figure 486). This area also produced most of the patterned daub. Much of the assemblage, 25% by weight, came from pits containing other redeposited material.

Figure 486: Distribution of daub by site area


Five fabrics were identified macroscopically, based mainly on the frequency and type of inclusions. Fabric B was the commonest, and has roller-stamp impressions and keying on the surfaces of 39545g of the pieces. Wattle impressions are present on pieces of Fabrics A, C and D, but the fragments in Fabrics E and F are almost featureless, or have poorly surviving detail.

The fabric descriptions are as follows:

  1. Fabric A A rather poorly mixed fabric with sparse sand, and fairly common vegetable tempering, often burnt out, giving a vesicular texture. The fabric is light in weight and typically cream to orange/red. 3079 fragments, wt 33.532kg
  2. Fabric B This fabric had fairly sparse sand and moderately sparse vegetable tempering; the colour was light brown to reddish. 6851 fragments, wt 242.967kg
  3. Fabric C This fabric had moderate sand and very sparse vegetable tempering. 2497 fragments, wt 30.616kg
  4. Fabric D Little sand or vegetable tempering, but some small fragments of pebbles. Rather a streaky mix of red and buff clays. There were few sherds in this fabric but since it only appeared in one context it may just be a badly mixed batch of clay. 499 fragments, wt 16.037kg
  5. Fabric E This fabric had much evenly sorted sand and small quartz grit but no vegetable tempering. 109 fragments, wt 1.112kg


The quantities of material deposited increase from the Late Iron Age into the Early Roman period, but the major period of deposition is mid-Roman (Period 4). Building 54 in Area G is the source of 75% of this peak. The rest of the mid-Roman daub appears to be the result of redeposition of earlier material.

Daub with wattle impressions

Figure 488: Daub, 3-8

The wattle impressions on the back of the daub have not, in general, been well preserved. Much of the material has been redeposited and is consequently abraded. This has meant that many of the diameters of the wattles were too incomplete to measure. Where it was possible to record measurements, they ranged from 12mm to 36mm in diameter. The wattles have generally been used with their bark still present (Figure 488, no. 3). There are also imprints of the squared timbers of wall studs, with the texture of the wood surviving on some pieces. The only measurable one was 64mm across (Figure 488, no. 4). The thickness of the clay mixture applied to the wattle framework varies widely from as little as 15mm to 45-50mm. None of the pieces have both inner and outer surfaces surviving together, so the total thickness of the wall cannot be estimated. Daub walls would naturally tend to split either side of the wattling. All the corners, both inward and outward facing, are fairly sharply moulded right-angles, and the flat surfaces have a wiped finish.

Daub with roller stamping and other surface treatment

Figure 489: Amounts of roller-stamped daub in dated contexts in Area G

There are 281 pieces of daub, weighing 32.263kg (9% of the site total, by weight), that are patterned with roller stamping (Figure 488 and Figure 489). For a discussion of this subject see Wickenden and Drury (1988, 86) and Russell (1990, 87-117). The patterns appearing on this daub are chevrons and diamonds, which are the commonest of the patterns found on Essex sites. The state of preservation and the relatively small fragment size means that it is not possible to say anything about the rollers used or the methods employed in executing the patterns. The distribution of sites with stamped daub focuses on Essex, London and North Kent (Russell 1990, 88) and Elms Farm adds another site without disturbing this pattern. Russell (1990, 107) discusses whether the patterning of daub was decorative or functional, concluding that it was more likely to be functional. The fragmentary nature of this material means that it does not add anything to his arguments. On the externally facing corners the stamping goes right up to the right-angled edge of the door or window frame, but is not present on the other face of the right-angle (Figure 488, no. 6).

The contexts around Building 54 in Area G (discussed in detail below) produced 96% of the roller-stamped daub from the site. Judging by the abraded and disturbed state of the fragments, they had been redeposited from the demolition of an earlier phase of this structure, or as backfill from elsewhere. The pottery from the contexts containing this distinctive daub was mid- to Late Roman. Dating the origins of the material is more difficult but the earliest evidence for the construction of Building 54 is late 2nd to early 3rd century.

The remaining 986g of roller-stamped daub came from a layer over Road 1 in Area F (10505, Period 4), oven 3042 in Area W (also Period 4), and machining layer 4000. Assuming that all the material is from a single building, it is difficult to explain why the material from Areas F and W was deposited so far away from the other roller-stamped daub.

Pit 15773 in Area M, layer 23105, and pit 23140 in Area N, produced eighty pieces of daub (Figure 488, nos 7-8) weighing 7282g, whose surfaces fragments have been impressed with chevrons and lattices made with a metal or thin wooden blade 60mm wide. Whether this is keying for rendering or a decorative feature is difficult to be certain, as it has only survived in relatively small fragments. A small quantity of unusual keyed daub broadly contemporary with the Elms Farm material was recovered from the Boudiccan destruction of Building 79 in the Culver Street excavations, Colchester (Crummy 1992b, 253-4). Most of the patterns are radiating 'scallop shell' shapes, though there were some lozenge patterns. Both these patterns are described as being scored but the Elms Farm examples appear to be impressed. Crummy suggests that the keyed daub is a primary feature of the building, and is likely to be military. This seems unlikely to be the case at Elms Farm. The keyed daub from Elms Farm occurs, as does the Colchester material, in Period 2B-3 features.

The daub from identified buildings:

Building 54 produced 45% of the daub from the site, by weight. The thirteen other buildings produced only 1900g of daub in total, most of which was featureless. The data are summarised in Table 90. The fact that there is so much more material from Building 54 than the rest of the site partly reflects the fact that the building burned down but, as discussed below, it is also due to burnt daub having been used as a packing material in the foundation trenches of the building.

Table 90: Daub from identified buildings
Building Count Wt (g) Period Fabrics Comments
1 1 4 2 A  
21 1 162 2 B  
31 5 22 2B A, B  
35 27 332 2B-3 A, B One fragment with a flat surface
38 1 56 3 B  
39 19 88 3 C  
40 2 10 3 B  
54 4192 162463 4-5 B Includes some roller stamping and wattle impressions
59 1 8 5 B  
61 19 276 4 B  
63 10 86 4 B, D  
66 14 412 4 A, B  
69 32 198 3 B, E  

The daub from Building 54

A total of 4192 fragments of daub in Fabric B, weighing 162.463kg, came from this area. Of this total, 24% (by weight) had roller stamping, 9% had flat surfaces and 28% had wattle impressions on the back. The wattle impressions range in diameter from 20mm to 36mm, where they were measurable. Four pieces have right-angled corners that are unstamped and face out, and may be the edges of doorways or window embrasures. Virtually all of the material from Building 54 (98%) comes from mid- to Late Roman features and the material in the remaining features appears to be residual.

Figure 142: Period 4 - Building 54 during excavation showing foundation slot 25219/25220

A sample of the material from foundation trench 7766 was lifted in a block, because the excavators thought it to be a burnt wall surviving in situ, but it proved to be disturbed fragments of burnt daub in earth, used as packing around the bases of burnt posts (Figure 142). Included in this loose material was a late 1st-early 2nd-century copy of a North Kent poppyhead beaker, burnt and lying on its side. The posts, which had been burnt in situ in the trench, were sufficiently charred to have lost much of their original shape, but two were measurable. One was 60mm across and 70mm long, and the other was c. 80mm across and 90mm long. The position of the buried parts of the posts might have contributed to their complete carbonisation - they were, in effect, in an anaerobic environment similar to that found in a charcoal maker's clamp. Collapsed material lying over the top of the posts would have helped to seal the trench. Once the bottoms of the posts had caught light, they could have smouldered for much longer than the parts of the posts above ground, while producing relatively little external heat, and having little effect on the surrounding earth. Neither the daub nor the earth packing in the trench showed signs of scorching, though the excavators noted severe heat damage to the north-west side of the wall trench.

While some of the other burnt daub from the area of the building may have derived from the fire that devastated the structure, the burnt daub from 7766 appears to have formed part of the original infill of the foundation trench, following the erection of the posts. This suggests that it was originally part of an earlier building, either an earlier phase of Building 54, or a different building altogether. The daub from the trench included roller-stamped daub, making it likely that all the roller-stamped daub from the building was redeposited. None of the other excavated buildings produced roller-stamped daub, and the building where it was originally used may have been outside the excavated area.

Daub and other baked clay from ovens and hearths

In total, 2414 fragments of daub and baked clay, weighing 13.012kg, were recovered from thirty-two of the excavated ovens and hearths on the site. The data are summarised in Table 91 below. The relatively small quantities in each of these features and the poor preservation of the material add little to the understanding of each structure. The wattle-impressed fragments may be evidence for the construction of the ovens, but might be intrusive building material. Although these structures range in date from Period 2 to 6 there is little visible difference in the nature of the material from them. The excavators considered a hearth to be any construction on or in which a fire is set, usually open, and an oven to be an enclosed heating structure, perhaps using heat indirectly. The differences become less well defined when the area of the oven is poorly preserved. Ovens and hearths containing more than 1kg of daub will be discussed in detail; otherwise, the material from them appears to relate as much to backfilling as use or construction. Baked clay from kilns is discussed in the kiln material report.

Table 91: Baked clay from ovens and hearths
(Fragments are featureless unless otherwise stated. Weights in grams)
Oven no. Group Area Period Count Wt Fabric Comments
9369 767 D 3 1 26 B  
9506 767 D 3 66 236 C  
9573 767 D 3 23 116 A, C  
10501 793 F 3 33 224 B  
6343 512 H 3 8 264 C Some flat surfaces
5692 3015 I 3B 2 18 B  
5789 628 I 3B 3 36 C  
5921 3015 I 3B 2 50 D  
13035 629 I 3B 47 116 C  
13187 627 I 3B 1 4 B  
13570 625 I 3B 4 4 A  
18638 625 I 3B 1 78 B  
4284 733 K 3-4 55 630 A  
4378 742 K 5 1180 6011 A, B, C Some wattle impressions and flat surfaces
4760 724 K 3 19 88 C  
4861 747 K 3 37 170 B Some flat surfaces
14491 703 L 3 26 388 B Some flat surfaces
14500 255 L 2C 1 30 B Wattle impressions
14550 5017 L 5-6 22 230 B  
14912 267 L 2B 2 132 B Flat surfaces
20090 721 L 5-6 17 368 A, E  
20137 721 L 6 4 258 A, C  
15984 695 M 4-5 251 686 A, C  
10894 205 N 2 5 32 B  
10895 205 N 2 21 470 B  
10929 205 N 2 3 92 B  
23157 686 N 3 4 88 C  
8509 213 P 2B 152 192 A  
8738 4029 P 4-6 267 1657 B, C Some flat surfaces
12364 974 R 6 25 48 C Flat surfaces
2338 929 W 4 104 117 B, C  
3014 916 W 4 28 156 C  
Total       2414 13015    

Catalogue of illustrated daub

Figure 488: Daub, 3-8

3. The well-preserved surface of the bark on a wattle impression. Fill 7536, Foundation trench 7752, Group 855, Building 54, Area G, Period 4

4. The imprint of part of a cut timber. Fill 7656, Group 865, Collapse Building 54, Area G, Period 4-5

5. Roller-stamped daub with wattle impressions on the reverse. Fill 7535, Foundation trench 7766, Group 856, Building 54, Area G, Period 4

6. A corner piece, with roller stamping on one face and a plain but rather worn surface on the other. This plain face is too abraded to be sure of the original surface type. Fill 7471, Group 865, Collapse Building 54, Area G, Period 4-5

7. Fragment with chevron keying. Fill 23142, Pit 23140, Group 681, Area N, Period 3

8. Fragment with chevron keying. Fill 23142, Pit 23140, Group 681, Area N

The Roman tile by Hilary Major and Ros Tyrrell

Cite this as: Major, H. and Tyrrell, R. 2015, The Roman tile, in M. Atkinson and S.J. Preston Heybridge: A Late Iron Age and Roman Settlement, Excavations at Elms Farm 1993-5, Internet Archaeology 40.


Table 92: Total amounts by tile type
  No. Wt (g) % by no. % by wt Average wt (g)
Tegula 21274 2615130 24.4 36.6 123
Imbrex 9722 695273 11.1 9.7 72
Brick 11445 2654435 13.1 37.1 232
Box flue 1034 116607 1.2 1.6 113
Spall 43883 1070705 50.2 15.0 24
Total 87358 7143424      

The excavation produced over 7 tonnes of Roman brick and tile, the largest assemblage of tile ever excavated from a rural site in Essex. There was little evidence on the site for the existence of domestic or other buildings utilising brick and tile in their construction, and the presence on the site of such a large amount of tile needs to be examined in this context. The total amounts by tile type are given in Table 92, and the percentages by number of pieces and weight, excluding spall, in Table 93.

Table 93: Percentages by tile type, excluding spall
  % by no. % by wt
Tegula 48.9 43.0
Imbrex 22.4 11.4
Brick 26.3 43.6
Box flue 2.4 1.9

Due to the sheer volume of tile, full details were only recorded for contexts with over 10kg, and for those contexts directly associated with them. All box flue tile was fully recorded, and the remainder of the tile was scanned. Statistical analysis was performed using sherd counts rather than weight, unless otherwise specified.

Details of the larger groups of tile, and tile associated with particular structures, can be found in the archive. The term 'brick' has been used for all flat tile.


Eleven fabrics were identified macroscopically during cataloguing. The principal inclusions present were chalk and sand in varying quantities. Chalk is a natural inclusion in the boulder clays of north Essex; while its presence will normally indicate that the tile was made from boulder clay; its absence does not necessarily imply that the clay used was not boulder clay. In addition, particularly with hypocaust tiles, burning or repeated heating may produce spurious changes of texture and colour in the tiles.

Most (80%) of the tile was in a single fabric group (Fabric A). This tile is likely to be local; it has sparse to medium inclusions of sand, and sometimes has sparse iron-rich flecks. The fabric is usually well fired and hard, and is usual orange-red to brownish-orange in colour, sometimes grey or with a grey core. The second most common fabric, Fabric F, formed 13% of the fully catalogued tile. This was a brown fabric, with a slightly soapy feel. It had fairly common sand, occasional blobs of paler clay, occasional iron-rich inclusions up to 3mm in diameter, and occasional very sparse small chalk flecks. The remaining fabrics occurred in only small quantities.

Eight fragments of tile in pale fabrics were present (Fabrics G and H). Tile in pale coloured fabrics occurs on many sites in Essex, and where it is original to a building it was probably used decoratively. On some sites, it is clearly reused, and this is likely to be the case at Elms Farm. Compared to some sites, there was very little 'gault' tile from Elms Farm; at Great Holts Farm, Boreham, for example, there were 138 fragments of 'gault' tile, probably all reused as building rubble (Major and Tyrrell 2003, 164). Although there were only eight pieces of 'gault' tile from Elms Farm, there were 414 pale-coloured tile tesserae, and the 'gault' tile may have been brought to the site as raw material for tessera manufacture. However, unlike the tesserae, which are highly concentrated in Area J, the 'gault' tile is spread across the site. There is therefore no overt connection between the 'gault' tesserae and the 'gault' tile.

Samples of Fabrics G and H were examined by I. Betts, who identified them as Molas fabric 2454 (Fabric G) and fabric 2455 (Fabric H). He writes: 'Molas fabric 2455 is believed to be an unusually fine version (lacking quartz inclusions) of fabric type 2454. It is believed to come from the same source as 2454, but may be from a clay band not normally used. The source of 2455 and 2454 is almost certainly north-west Kent, probably in the area of Eccles villa (although there may be other tileries in the area exploiting the same clay source).' Other sites in the area with Eccles tile include Colchester (Betts 1992), where it was noted that tile in this fabric was probably not being imported into the area after the 1st century.

Not all the pale-coloured tile found in the region is necessarily from outside Essex; a few of the pieces from Great Holts Farm, Boreham, may have been from the Colchester area, having a fabric extremely similar to Colchester white ware pottery (P. Sealey, pers. comm.; Major and Tyrrell 2003, 164).

A certain amount of overfired and distorted tile was found, probably kiln wasters, although this does not imply that there were tile kilns on the site. While they could not be used as roof tile or visible building material, they were still useful as building rubble. At the bath complex at Great Holts Farm, Boreham, for example, the presence of mortar on wasters demonstrated that they had been used in the fabric of the building (Major and Tyrrell 2003). There were 350 pieces of overfired brick from Elms Farm, some virtually vitrified, 211 pieces of tegula and a minor amount of other tile.

Tile distribution by period

Tile was recovered from 2287 contexts, with an average of 38 pieces per context, weighing 3127g. Table 94 gives the overall figures for contexts that are closely dated. Period 5-6 has been included as a 'closely dated' period for a number of reasons. Most of the phased contexts covering more than one period had relatively little tile, but Period 5-6 has the second greatest amount of tile for any period, and to exclude it would mean that the amount of tile deposited in the later and latest Roman periods would not be truly represented in the analysis, nor would the amount of tile from Area J, since contexts dated to Period 5-6 are concentrated in that area. Nearly half of the tile from Area J comes from Period 5-6.

Table 94: Tile numbers and weights by period (including spall): discrete periods only
Period No. of pieces Wt (g) Average wt per sherd (g) % by number % by weight
Prehist. 5 54 11 0.0 0.0
2 777 48609 63 1.3 0.9
3 7684 503778 66 13.0 9.7
4 17409 1477620 85 29.5 28.3
5 8765 939713 107 14.9 18.0
5-6 12076 1229952 102 20.5 23.6
6 11442 980644 86 19.4 18.8
7 849 38262 45 1.4 0.7
Total 57924 5218632      

It can be seen from Table 94 that the mid-Roman period had the greatest amount of tile deposition. There is an apparent fall-off in deposition in Period 5, although if the tile from Period 5-6 is split between the two periods, it is not a very great fall-off. The average sherd size gradually increases up to a maximum of 107g in Period 5, falling off slightly in Period 6. For Period 7, for which all of the Roman tile must be residual, average sherd size has dropped considerably, to only 45g.

With a site such as Elms Farm, where not all features were excavated, it is difficult ascertaining from data such as that in Table 94 whether the apparent increase in tile deposition is a result merely of more features from a certain period having been excavated. Period 2/3 features may be under-represented, for example, because they lay at the bottom of vertical stratigraphy where this was present. The percentage of excavated contexts containing tile was calculated for each period. The graph (Figure 490) clearly shows that the proportion of contexts containing tile increases with time up to Period 5, and decreases only slightly thereafter.

Figure 490: Percentage of contexts with tile

All the Roman tile from Period 7 must be residual, and the pattern of deposition for this purely residual group can be compared with the earlier periods. As noted above, average sherd size in this period was considerably lower than in the preceding periods. This is at least partly due to there being relatively little brick in Period 7 contexts (only 4%, compared with 11-16% from Periods 2-6). As the average piece of brick weighs more than the average piece of roof tile, the lack of brick will tend to make the average sherd weight lower. Contrary to expectations, the residual tile in Period 7 contexts does not include a higher proportion of spall than the earlier contexts; 50% is spall, which is a slightly lower proportion than Periods 2-6, and not much higher than Periods 5-6. Nor, as can be seen from Figure 490, is the proportion of contexts containing tile greatly different from the preceding periods. The average weight of tile per context is, however, much lower than in the Roman contexts (Figure 491), at only 649g. In contrast, Period 5-6 has an average of over 10kg of tile per context. In summary then, the residual tile from Period 7 is almost as widespread as the tile from the preceding periods, and does not contain a significantly higher proportion of spall. There is, however, far less tile, on average, from each context, with a smaller average sherd, probably due to the relative scarcity of brick fragments.

Figure 491: Average weight of tile per context (discrete periods only)

In Period 2 (i.e. potentially pre-Roman) 157 contexts produced Roman tile, including twenty-one Period 2A contexts, which should be entirely pre-Roman. The majority of the contexts contained only one or two sherds, which could be construed as intrusive. There are some contexts with much larger amounts, however, which in most cases probably represents continued deposition in the Early Roman period.

The spatial distribution of the tile

The distribution of the tile by area is given in Table 95, showing a concentration around the temple (Area J), and in the surrounding areas. In general, the outlying areas of the site have smaller amounts of tile present. No analysis of the distribution by area/period was undertaken owing to constraints of time.

Table 95: Gross amounts of tile by area (number of pieces)
Area Total %
D 3607 4.2
E 4225 5.0
F 7675 9.0
G 4735 5.6
H 12599 14.8
I 4162 4.9
J 17658 20.7
K 7343 8.6
L 3061 3.6
M 6424 7.5
N 2606 3.1
P 3233 3.8
Q 171 0.2
R 5996 7.0
W 1639 1.9
Total 85134 -

Table 96 shows the distribution by feature type. Over half of the tile came from either layers or pits (including wells). The tile from the pits and ditches was probably deposited as 'rubbish', although a certain proportion of this material is likely to be residual, particularly in areas where features cut earlier features. In the case of features such as cremations, it can be assumed that all the tile is residual, or at least not a deliberate dump of rubbish. On some sites tile can be directly associated with a burial, being used as a lid for a cremation urn, or to build a cist, but there is no convincing evidence for such practices at Elms Farm.

Although there is no evidence for the use of tile for roofs on the site, it was used in the construction of some of the hearths and ovens, and in other structures such as walls. Tile that had definitely been used in constructions accounted for only 14% of the tile from the site.

Table 96: Distribution by feature type
Group No. %
Layers 32655 37.4
Pit/well 24274 27.8
Ditch/gully 8947 10.2
Building contexts 7824 9.0
Hearth/oven/kiln 4334 5.0
Machining layers 3072 3.5
Uncertain 2953 3.4
Demolition 2193 2.5
Surface 706 0.8
Cremation/grave 317 0.4
Other 76 0.1
Natural 7 0.0
Total 87358 -

Tile with salt splashes

The presence of splashes of salt glaze on the tile was recorded in the hope of helping to identify areas of the site where secondary salt processing, or other processes involving salt, might have taken place. Splashes of salt glaze on tile can be produced accidentally when a tile with salt on it is heated, and might be expected to occur, for example, on a tile hearth used for heating brine. It is possible that some of the salt glaze splashes were created at the tilery, but where they occur on the broken edge of a tile, this is less likely, and it seems reasonable to assume that the glaze was the result of an incident on the excavated site.

Such splashes proved to be fairly rare. There were seventy-one fragments with salt splashes on them, from sixty-two contexts. The type of tile with salt splashes appears to be of significance, as 63% are pieces of brick, which forms only 13% of the assemblage as a whole. Brick is the most suitable type of tile for the building of hearths and ovens, and four of the pieces of brick from oven construction 15984 had salt splashes. However, none of the other pieces of brick were from ovens or hearths, and only two other such features had any salt splashes, both single splashes (Hearth 9573 and Oven 20137). It is possible that the bricks with salt splashes came from demolished hearths, but this is difficult to prove.

The distribution of tiles with salt splashes is quite different to that of the tile as a whole. The areas with the largest number of pieces were M, N, D and H, with only four pieces from Area J, which yielded nearly 20% of the total amount of tile.

Figure 492: Comparative distribution of briquetage, and tile with salt splashes (Periods 3-7)

Given that the site has an abundance of material definitely connected with salt processing, i.e. the briquetage, the distribution of the salt splashes was compared with that of the briquetage. Briquetage from Period 2 was excluded from the study, as it is unlikely that material of this date would have been used in association with tile (although a few of the salt splashes are from Period 2 contexts). The distribution shows broadly similar trends (Figure 492), though the peaks for each material occur in different areas of the site, apart from Area D, which has relatively large amounts of both briquetage and salt splashes. The two main peaks are, in both cases, in adjacent areas of the site (G/H and M/N), which is curious, but not necessarily significant. There is also a minor briquetage peak in Area K, which has no tile with salt splashes at all. Twenty-three contexts, predominantly pits and layers, contained both tile with salt splashes, and salt briquetage, but none of these appear to be significant.

Figure 493: Comparison of the chronological distributions of tile, tile with salt splashes, and salt briquetage. Each period is shown as a percentage of the total dated assemblage

However, when the chronological distribution of the salt briquetage and the tile with salt splashes is compared, a different picture emerges. Figure 493 compares these two categories with the tile as a whole, and it is evident that the salt-splashed tile has a distribution that is very similar to the tile (bar Period 5-6), and dissimilar to the salt briquetage, which is heavily concentrated in Periods 2 and 3. It therefore seems unlikely that the briquetage and the tile with salt splashes were used in conjunction with one another.

In conclusion, then, it is likely that some, if not all, of the tile with salt splashes represents the residue of on-site processes involving salt and heat, as demonstrated by the evidence from Oven 15984. Most of the material is not from primary contexts, but may have derived from hearths and ovens in the vicinity. There is no strong connection between the distribution of salt briquetage, and the distribution of the salt-splashed tile, and they were unlikely to have been used together.


Only three tegulae from the site had both a length and a width present, but seventeen others had either the width or the length measurable. The range of lengths was 375-430mm, and widths 290-375mm. By comparison, at Great Holts Farm, Boreham, the range of lengths was 350-490mm, and widths 245-364mm (Major and Tyrrell 2003). The three complete tiles measured 410x300mm, 420x310mm, and 378x310mm. Contrary to expectations, few of the measurable tegulae came from structures such as ovens.

Flange and cutaway shapes were only recorded for the fully catalogued tile. Details of the analysis of these features may be found in the archive report. The basic flange shapes present of the 2068 tegula flanges examined are illustrated in Figure 719. The flange shapes may be divided into two basic groups, rounded flanges (types 1, 2, 5, 7, 8, 10, 11, 12 and 13), and squared flanges (types 3, 4, 6 and 9), often quite sharply cut. It is accepted that flange profiles may vary somewhat on a single tile, but any variations will normally be within the same group. Types 3 (44% of the total number of flanges), 1 (25%) and 4 (15%) were most common for all fabrics, and for all periods of the site.

Figure 719: Roman brick and tile - flanges and cut-aways

The cutaway types present are illustrated in Figure 719. The lower end cutaways were designated A, and the upper end B. There does not appear to be any significant chronological distribution of cutaway types at Elms Farm. Seven tegulae had cutaways present at both ends, six of them having A1 and B7, and the seventh A1 and B5.

Seven lower end cutaway types were present. Only 221 of the 353 examples noted could be assigned to a specific type. Type A1 was the most common cutaway, present on 75% of the tegulae examined. Brodribb (1987, 17) cites his type 1, equivalent to A2 here, as the most common cutaway type country-wide, occurring on some 75% of all tegulae. Even including type A3 as a variation of type A2, together they form only 5% of the cutaways at Elms Farm. Some comparisons may be made with other sites in Essex. At most sites A1 appears to be most common, as here, and A2 scarce. However, the cutaway here designated A8 was by far the most common at Chelmsford sites S and AR (Wickenden and Drury 1988, 80), but is almost completely absent from Elms Farm, which produced only a single example.

At the upper end of the tile, seven different cutaway types were noted. Of the 248 upper cutaways recorded, 137 could be assigned to a type. Type B7 was most common, with 105 examples (77%). The upper end cutaways are simpler to make than the lower end, normally involving only two knife cuts, and it is possible that most of the different types could merely be misshapen examples of B7. The exception is B4, with its single, curving cut, of which there are two examples from Elms Farm.

Seventy tegulae had nail holes, fifty-seven of which were circular, two oval, four square, one rectangular and one irregular. The rest were not complete enough to be certain of the shape. Most of the nail holes were made by pushing an implement through the soft clay of the tile before firing, normally from the upper surface, although on one example the hole had clearly been made from the bottom of the tile. In four cases, the holes had been drilled after firing. The diameters of the circular holes were between 3mm and 19mm, the average diameter being 10mm. Since tegulae were probably not being used for roofing at Elms Farm, the occurrence of nail holes here is purely incidental.

Signatures on tegulae

There are 969 pieces of tegula with signatures, comprising a combination of arcs and lines. Most of the signatures are partial, and only 182 tiles have definitely complete signatures.

Arcs were far more common than linear signatures, forming 96% of the overall total of 969 signatures. The most common form of signature consisted of three arcs placed against the edge of the tile (fifty-four examples, 30% of the total), followed by two arcs placed against the edge (forty-four examples, 24%). The bulk of the signatures were on tiles in fabric A, or on the scanned tegulae, whose fabric was not recorded. It is therefore not possible to comment meaningfully on any differences in signature between the fabrics. No particular pattern could be seen in the chronological distribution of the signatures.

Only the more unusual signatures have been illustrated here, as these are the ones which, in the long term, are most likely to suggest intra-site links.

Figure 494: Signatures on tegulae, 9-17

9. Wavy line adjacent to the edge. Layer 4689, Area K, Period 4-5. Similar fragments came from Fill 10891, Pit 10910, Group 676, Area N, Period 5, and Layer 24221, Area M, Period 4-5

10. Double arc cut by a single line. Layer 9427, Area D, not phased. This signature can be broadly paralleled at a number of other sites in Essex and elsewhere. Single arcs crossed through with a line appear to be fairly common; they include examples from Beauport Park (Brodribb 1979, fig. 5.13), Southwark (Crowley 1992, fig. 47.19), Castleford (Betts 1998, fig. 99.1) and Great Holts Farm. Double arcs cut by lines are rarer, and generally are cut by two lines, as with examples from Southwark (op. cit. fig. 47.20) and Great Holts Farm.

11. Two intersecting arcs against the edge. Fill 10061, Pit 10062, Area E, Period 4. Three other pieces have similar arcs, some showing more detail of the second arc; they are from 16184 (Ditch 16185, Group 584, Area H, Period 6), 20194 (Pit 20193, Group 718, Area L, Period 5) and Layer 21668 (Group 446, Area J, Period 6)

12. Four intersecting arcs; a very incomplete signature, but possibly paralleling the two intersecting arcs from 10061. Fill 12246, Well 12245, Group 955, Area R, Period 3-6

13. Three arcs away from the edge. Fill 15043, Pit 15042, Group 650, Area M, Period 6

14. Three arcs. Fill 21883, Post-hole 21881, Group 5009, Area J, Period 5-6. A number of other fragments have three arcs, positioned away from the edge of the tile, including pieces from 10688 (Group 842, Area F, Period 5) and 21823 (Group 5009, Area J, Period 5-6), although they are not similar enough to suggest that they are from the same maker.

15. Three incomplete arcs in a tear-drop shape. Fill 15050, Ditch 15049, Group 473, Area M, Period 6

16. A single curved line springing from one corner of the tile. Although it is unusual to find a signature at the upper end of a tegula, the length of the tile is incomplete, and this is probably just a grandiose flourish from one end of the tile to the other. Cleaning layer 15461, Area M, Period 5-6

17. Two arcs, probably forming an oval against the edge of the tile. Fill 24398, Well 22210, Group 448, Area J, Period 6


The majority of the imbrices were unremarkable. One feature noted as being of possible interest was a distinctive flat band along the crest of some of the imbrices from 21823 (Period 5-6). It was hoped that this might be a feature that could be used to look at site formation processes, but only one other context proved to contain a similar tile, layer 21668 (Period 6), which is in the same vicinity. Only a few sherds had markings, including one piece with a signature-like mark, and one with a graffito (no. 54, below). A small number of imbrices had measurable dimensions, tabulated in the archive report.

The following sherds had features of note (not illustrated).

18. Edge, with two shallow V-shaped notches 35mm apart. These were possibly caused by resting the undried tile on sticks. The tile is very abraded. Fill 50, Foundation pit 15, Group 932, Area W, Period 6

19. Corner, with part of a finger-drawn arc similar to tegula signatures, probably forming a quarter arc across the quarter. Another fragment with two arcs (4657) is more likely to be from a very thin tegula rather than an imbrex. Machining layer 4000

20. Imbrex fragment, probably deliberately chipped into an oval disc. Fill 15076, Ditch 15049, Group 473, Area M, Period 6

21. Two imbrex fragments shaped into rectangles, 94x47x14mm. These are presumably from flooring. Layer 16187, Group 573, Area H, Period 5-6

Box flue tile

There were 1034 fragments of box flue and voussoir tile, forming 1.2% of the tile from the site. By comparison, the excavation of Chelmsford bath-house (site CF20; Major in prep.) yielded only 541 fragments, though they were larger on average, 218g compared to an average of 113g at Elms Farm. The total weight of box flue tile from Chelmsford bath-house was 118kg; Elms Farm produced 107kg. There were 752 pieces of box flue tile from Great Holts Farm, Boreham, still only 1.6% of the total tile, but forming a much higher proportion of the tile in the vicinity of the bath-house (Major and Tyrrell 2003, 170).

The voussoirs will be considered separately from the box flue tile.

Only nineteen box flue tiles had measurable dimensions, and only one had all three dimensions present. The latter (illustrated as combing patterns 11 and 14) was 410mm high, 155mm wide, and 110mm deep. There were no other measurable heights. Widths varied from c. 125mm to 170mm, and depth from 80mm to 160mm. By comparison, Brodribb (1987, 143) gives the average size of a box flue tile as 366x190x131mm. The complete box flue tile from Elms Farm is thus taller and narrower than the average, and the widths overall consistently less than average.

Ninety-three fragments had parts of the cut-outs in the side of the tile. Fourteen of these were voussoir tiles; with one exception all had circular cut-outs. There were no complete cut-outs, apart from those on the voussoirs, and the cut-outs on the box flue tiles could only be categorised as having either curved edges (from circular or oval cut-outs) or straight edges (from rectangular or, less likely, square cut-outs), apart from a single example of a rectangular cut-out, and part of an irregular quadrilateral, probably a poorly cut rectangle. Straight edges were slightly more common than curved edges, forming 55% of the total number of cut-outs. No attempt was made to correlate cut-out shape with combing or cut lattice decoration, as only four of the cut-outs were from lattice cut tiles.

Two styles of surface treatment were present; incised cross-hatching and combing. Combing was present on 797 tiles and 52 had cut lattice decoration. Incised cross-hatching occurred on 52 fragments, which represents less than 7% of the decorated box flue tile. No detailed analysis of the pattern was carried out. Although the technique was present on five different fabrics, it was most frequently found on tiles in fabric F, where it represented nearly 14% of the decorated tile in that fabric, as opposed to 4% for fabric A.

It has been suggested (Black 1985) that incised cross-hatching is an early Roman type of design, and cut lattices are certainly particularly common on Early Roman tile. At the Marlowe Car Park site in Canterbury, for example, all the box flue tile definitely from Period 2 (c. AD 70/8-100/110) had cut lattice keying (Black 1995, 1268-9). At Elms Farm cut lattice tile forms a higher proportion of the examples from Periods 2 and 3 than it does in later periods, tending to support the hypothesis that it is an Early Roman type. There were no contexts containing large groups of lattice-cut tile, and none of the pieces have mortar on the broken edges, which would suggest that they were not reused as building rubble.

The combed decoration present was recorded in detail, using a scheme for cataloguing individual elements. This enabled the smaller fragments of box flue tile to be related to those pieces with more complete patterns. The number of teeth in the combs used was not recorded as a matter of course. This method of recording individual elements was first used for the box flue tile from two Chelmsford sites (CF16 and CF20) in 1991, and was used in the analysis of all tile assemblages excavated by Essex County Council Field Unit until 2003, enabling comparisons to be made between different assemblages.

Ninety-nine different pattern elements were identified, excluding examples with complex combing, which were recorded individually. The assemblage from Chelmsford bath-house included fifty-nine different elements, suggesting a wider range of sources for the Elms Farm box flue tile. Combing was present on tile in fabrics A, B, C, D and F, although most of the combed tiles were in fabric A (a total of 611), and the illustrated tiles are all in this fabric. There were 129 combed tiles in fabric F, with very minor amounts in the other fabrics.

Fifteen patterns were identified as being definitely, or almost definitely, present, as follows.

Straight line patterns

  1. Saltire with full frame
  2. Saltire with side lines
  3. Saltire with central and side lines
  4. Saltire, no frame
  5. Multiple vertical lines
  6. Full frame, with or without divider
  7. Frame, diagonal line
  8. Vertical lines, V over
  9. Vertical line, short oblique lines

Wavy line patterns

  1. Multiple vertical waves
  2. Multiple horizontal waves
  3. Vertical line/wave/line
  4. Vertical wave/line/wave
  5. Vertical wave with frame
  6. Horizontal wave/line/wave

Straight line patterns were far commoner than wavy lines in fabric A, occurring on 97% of the fragments. Straight line patterns were also most common in fabric F, but to a lesser extent (77%). The straight line patterns that were potentially the commonest were numbers 8, 5, 9, and 2, in that order. The most common wavy line patterns were 12, 13, and 14. The presence of pattern 7 is somewhat tentative; the largest fragment (illustrated) could possibly be part of a saltire cross with a frame and divider.

In addition, there were a small number of unassociated elements that did not fit into the above scheme, none of which were complete enough to be confident about suggesting an overall pattern. These include a possible saltire with a top line, a saltire cross with a horizontal line, and a 'Union Jack' pattern.

Figure 495: Box flue tile, 22-40

22. Pattern 1. Layer 6008, Group 590, Area H, Period 5-6

23. Pattern 2. Fill 10555, Pit 10580, Group 828, Area F, Period 4

24. Pattern 3. Fill 16230, Ditch 16231, Group 584, Area H, Period 6

25. Pattern 4. Fill 10555, Pit 10580, Group 828, Area F, Period 4

26. Pattern 5. Fill 14558, Pit 14529, Group 722, Area L, Period 6

27. Pattern 6. Fill 10555, Pit 10580, Group 828, Area F, Period 4

28. Pattern 7. Spread 6195, Group 576, Area H, Period 4-5

29. Pattern 8. Context 2388, Oven construction 2647, Group 929, Area W, Period 4

30. Pattern 9. (a) Layer 5494, Area I, Period 3B (b) Fill 20815, Ditch 20814, Group 466, Area L, Period 6

31. Pattern 10. Fill 10555, Pit 10580, Group 828, Area F, Period 4

32. Pattern 11. Multiple horizontal waves: not illustrated

33. Pattern 12. Fill 10555, Pit 10580, Group 828, Area F, Period 4

34. Pattern 13. Fill 10555, Pit 10580, Group 828, Area F, Period 4

35. Pattern 14. Context 2337, Oven construction 2647, Group 929, Area W, Period 4

36. Pattern 15. Machining layer 4000

Elements not assigned to a pattern

37. Context 8591, Construction cut 8656, Group 3052, Area P, Period 3

38. Cleaning layer 9001, Area D, not phased (element B13)

39. Fill 10296, Ditch 10406, Group 838, Area F, Period 5-6

40. Layer 12041, Group 966, Area R, Period 4

Voussoir tiles

Thirty-four pieces of definite or possible voussoir were identified. Possible voussoir fragments were those with combing on adjacent sides, with cut-outs on the combed faces, or with a depression round a curved cut-out, a feature that was present on some of the definite voussoirs. Some pieces of voussoir were very similar to each other, and were undoubtedly part of same batch. Five groups (A-E) were identified based on stylistic similarities, and, in the case of group A, the use of the same tile comb. Ten fragments could not be fitted into the groups. Most of the voussoirs were from Period 5-6, although they occurred in Period 4 contexts onward. The largest group came from post-hole 21881, context 21883 (Area J, Period 5-6), where they formed part of the post packing.

Group A

Four tiles from context 21883 were assigned to this group. The cut-out face has a circular hole with a diameter of 27-29mm. The inner edge of the hole appears to have been formed by a knob on the mould, as it is sanded, with the outer edge having been cut while the voussoir was still on the mould. The implement used to cut the hole has left a circular 'step' round the hole, 65mm in diameter, visible on two out of the three pieces of cut-out side present. One tile has a complete height of 182mm.

The cut-out face has a pattern made from four strokes of the comb surrounding the hole; the face without the cut-out has a three-stroke pattern (two examples present). The comb used had seven teeth, with a distinctive gap between two of the teeth. Single fragments decorated by the same comb came from eleven different contexts across the site.

Figure 496: Voussoir tiles, 41-44 (all context 21883)

41. Almost complete face with cut-out, showing the step around the hole.

42. Joining pieces forming one-and-a-half sides.

43. Almost complete face with a small part of the adjoining side.

Group B

Three sherds from context 21883 were assigned to this group. The combing pattern is a saltire cross with side lines, and a vertical line drawn in two separate strokes. It was done with an eight-toothed comb, the end tooth not always visible. The combing occurs on faces with or without cut-outs. The fabric is different from group A, with a distinctly grittier feel. There is only one other definite piece of group B voussoir from a context other than 21883 (Pit 15206, Area M, Period 5), but there is a possible fragment from Layer 10104 (Area F, Period 4), and a half face with a full height of 175mm from Layer 5275 (Area J, Period 5-6) which is probably group B. The latter tile was not definitely decorated with the same comb as the other pieces in this group. It possibly lacks the vertical line, and one diagonal is two strokes wide.

44. Substantially complete face. Context 21883

Group C

Faces with circular cut-outs, but without combing; two pieces from context 21883. The fabric is as group B. One piece has a small area of combing present on the adjacent side, but it is not the same pattern as group B.

Group D

One sherd, with a circular hole with a ledge round it, as Group A, but on an uncombed face (Cleaning layer 5543, Area J, Period 5-6). A small part of the adjacent combed face is present, comprising a frame round three edges. A fragment from cleaning layer 5662 (Area I, not phased) has a very similar hole, and is possibly part of the same tile.

Group E

One sherd from Slot 4695, Area K (not dated). A combed face with a circular hole with a ledge round it, as group A. The pattern was probably similar to group A, but a different comb was used, with six, rather than seven, teeth. Another sherd (Pit 20193, Area L, Period 5) was probably combed with the same comb.

Bessales and pedales

Bessales and pedales are square tiles, mainly used to make the under-floor pillars for a hypocaust system. Bessales are normally about 200mm square, and pedales are about 280mm square. There were four complete bessales, measuring 215x207mm, 205x203mm, 205x200mm, and 200x190mm, 200x197mm and 198x197mm, and five other incomplete fragments that must be from bessales. Two of the complete bessales had been used as a post-pad, and one came from an oven structure. A fourth was from a pit, and there were two unstratified tiles. There was a single complete pedalis, 302x300mm, from an oven structure. Incomplete pedales are difficult to identify, as they have the same range of width and thickness as lydion.

Sesquipedales and bipedales

There was one probable example of a sesquipedalis (a nominal one-and-a-half Roman feet square) from corn drier 15984 (context 15662), 400mm wide, at least 350mm long and 33mm thick. Bipedales (two Roman feet square) are not definitely present, although twenty-five fragments with thicknesses of 50mm or more may have derived from them.


Lydion were commonly used as wall bonding tiles, and had a nominal size of one by one-and-a-half Roman feet. Nineteen whole lydion were found, ranging in size from 375x265mm to 410x290mm and in thickness from 27 to 37mm. Forty-four other tiles were broken in such a way that their proportions suggested that they belonged to this category. They ranged in width/length from 270mm to 405mm. The majority of the more complete lydion from this site were associated with structures, either from masonry or ovens.

Eight pieces of lydion from 21997 (two courses of tile from repairs made to the precinct wall in Period 5) have a depression, probably a thumb print, in the middle of one end, c. 50-70mm in from the edge. In one case, the depression is somewhat off-centre, but there is a stone in the spot where the hole would have been if it were centrally placed. None of the pieces included both ends of the tile, so it is not known whether there was a hole at both ends. The marks may have been made when the tile was picked up out of the mould, although there are no traces of fingerprints on the bottom. Holes like these were not recorded from elsewhere on the site, although they would not be noticeable in isolation since they look like holes where stones have fallen out of the tile. The only reason these were spotted is because they form a large group from a single context. All are in the same fabric. There is also one piece from the same context with four fingerprints on the edge of the tile. The position of the prints suggests that the tile was being held in a vertical position at the time.

Figure 497: Lydion and Opus spicatum, 45-46

45. Lydion with hole at one end. Context 21997, Group 439 , Period 5

Brick with ledges

There were 122 pieces of brick from the site with shallow ledges along one or more edges. The ledges appear to have been formed during manufacture, probably by strips of wood being pressed onto the edge of the tile. Where they occur on adjacent sides, they overlap at the corner, and could not have been created when the bricks were stacked for drying. They appear to have no practical function, and could just be an artefact of the method of manufacture. Four of these bricks had measurable dimensions, two being bessales from 18924 (205x203mm and 205x200mm), one a lydion (450x298mm, from 6754), and one either a lydion or a pedalis (W. 280mm, from 20137). The range of thickness for all pieces was 27-44mm. This feature thus occurs on a range of different types of tile, and also in several fabrics. This suggests that they may have come from more than one tilery.

The distribution by period is of some interest. Ledged bricks were found in contexts of Period 3 onwards (the single fragment from a Period 2 context is presumably intrusive), but the largest proportion comes from Period 4 contexts. Thirty-six of the ledged pieces from Period 4 came from a single feature, Oven 4284, where they had been used in the structure. Deposition of ledged bricks falls off markedly after Period 4. The evidence therefore suggests that ledges were an Early and mid-Roman feature. Similar 'margined' tiles were noted from Chelmsford Temple (Wickenden 1992, 63), a total of nine pieces, the earliest coming from a 2nd-century context.

Opus spicatum

One fragment of narrow brick was found. It is presumably from an opus spicatum brick, used for herringbone floors, although whether it was actually used in this way on this site is debatable. Opus spicatum bricks are relatively rare. Local examples include finds from Colchester (Crummy 1992b, 256) and some from the temple precinct at Great Chesterford, examined by the writer in Saffron Walden Museum. The latter bricks share with the Elms Farm brick the characteristic of having one cut edge, the other edges being sanded.

46. Segment from a brick in fabric A. It has a complete width of 86mm and is 37mm thick. One edge is cut, the other edge and the back are sanded. Both ends are broken. The cut edge possibly has the end of an arc against it. One end has broken across a recess cut in the back, with a rectangular section. Context 20263, Pit 20193, Group 718, Area L, Period 5 (Figure 497)

Signatures on bricks

There were 564 pieces of brick with 'signatures' on them, 9.5% of the total. Most of these were probably lydion. Three basic groups of signatures were defined; those consisting of one or more arcs, those consisting of lines, and complex signatures. The most common signatures consisted of one, two or three arcs, normally placed against the end of the tile where this was present. The bricks with signatures were principally in fabric A, with only minor amounts of other fabrics represented.

The distribution of the signature groups was considered by period, to examine whether there was a discernible chronological difference in the use of arcs or lines. The proportion of linear signatures to arcs changed little over time, and it likely that each type of signature was used throughout the Roman period. A few of the more complex signatures are illustrated.

Figure 498: Signatures on bricks, 47-50

47. Two sets of intersecting lines against the edge. 4140, Pit 4139, Area K, Period 6. Similar signatures came from 6316, Layer, Area H, Period 4; and 8000, cleaning layer, Area E, not phased. Single intersecting lines came from 4020, oven 4021, Group 728, Area K, Period 3

48. Two arcs, probably an oval, not against the edge, cut by a line. Cleaning layer 6515, Area H, not phased

49. Complex lines and a ?circle, possibly a graffito rather than a signature. The piece of tile is shaped like a fish, perhaps deliberately, with the lines forming the backbone and ribs. Fill 12059, Ditch 12046, Group 969, Area R, Period 4-6

50. Two arcs and a curvilinear line. Fill 13740, Foundation trench 13741, Group 170, Area J, Period 2B-3


A number of pieces had graffiti on them, but most were frustratingly incomplete, consisting of mere lines against broken edges. None seemed obviously parts of literate graffiti. Most graffiti were made pre-firing, and relate to activities at the tilery rather than the site on which they are found. One piece of imbrex (no. 54) may be an exception, as the marks were made after firing. The more complete graffiti are illustrated.

Figure 499: Graffiti, 51-55

51. Tegula. A series of lines and arcs made with a stick, including an arc against the edge which is analogous to a signature. Fill 4386, Pit 4153, Group 744, Area K, Period 6.

52. Tegula. A number of short, sinuous lines, roughly parallel, with a short transverse line cutting them, and other short lines elsewhere on the sherd. This does not appear to be a literate graffito, but could be imitating writing. Fill 8167, Well 8188, Group 788, Area E, Period 3.

53. Tegula spall. A small fragment covered with short curvilinear and straight marks made with a stick. Fill 9028, Pit 9029, Group 783, Area D, Period 3B

54. Imbrex. Marks made by a stick or similar object, evidently after firing, since they cut the darker surface. They are definitely not modern, as they are weathered. This is probably part of a crude representational drawing, though the subject is obscure. Fill 18742, Robber trench 22169, Group 447, Area J, Period 6.

55. Brick. Fragment with four cut lines. Fill 16262, Pit 16263, Group 566, Area H, Period 5

Animal and human prints, and other impressions

Animal or human prints were present on 119 tiles. Animal prints were quite rare on this site; 0.14% of the tile (by number) had animal or bootprints, which may be compared with the site at Bull's Lodge, Boreham, just outside Chelmsford (Major 1993), where 0.36% of the tile had such markings, and CF20, with 0.2%. The quality of the evidence was rather disappointing, given the overall amount of tile, with a high proportion of incomplete prints. There were, however, some potentially unusual prints present, including possible rodent prints, a probable bird, and a probable piglet. Overall, dog prints were most common, followed by cats. Since the tile from this site must have derived from a number of different tileries, analysis of the prints is not useful for determining the environment of a particular tilery. The prints present are summarised in Table 97. There were also a number of other impressions present, some of which are illustrated.

Table 97: Animal and human prints present
Type of print No.
Bird 1
Cat 16
Cat/dog 2
Cat? 9
Deer? 3
Dog 40
Dog or fox 1
Goat? 2
Hobnails 14
Hobnails? 1
Hoof 4
Human foot 1
Human foot/hand 1
Human foot/hand? 1
Human foot? 2
Human hand 2
Human hand? 2
Pig 1
Piglet? 1
Rodent/bird scratches 1
Roe deer? 1
Sheep/goat 5
Sheep/goat? 1
Sheep? 1
Unidentified 6
Total 119

Animal and human prints

Figure 500: Animal and human prints, 56-59

56. Partial shoe print on brick, comprising six hobnails diam. 6-8mm, with four much smaller circular impressions set round the edge. Prepared surface 16181, Group 569, Area H, Period 6

57. Not illustrated. Partial shoe print on a tegula, with six hobnails. Five are circular or slightly oval, diam. c. 6mm; the sixth, unusually, is rectangular, measuring 8x5mm. 10688, Pit 10763, Group 842, Area F, Period 5

58. Incomplete impression of a petalled stud, on an imbrex. This was probably made by a copper-alloy stud. Fill 13216, Pit 13358, Group 645, Area I, Period 5

59. Impression of a hollow object with a square section, possibly a cut bone, or the end of a tool handle. Fill 13277, Pit 13358, Group 645, Area I, Period 5

60. Not illustrated. Brick, with a circular hole with a rounded bottom, diam. 15mm, 15mm deep, and 18mm in from the edge of the brick. This impression does not appear to have been made by a finger, but possibly by the butt end of a tool handle. While ostensibly similar to the holes from 21997, it is definitely not part of that batch of tile. Fill 9253, Pit 9243, Group 806, Area D, Period 4

61. Not illustrated. Tegula. At least two overlying sets of thin, parallel scratch marks, possibly made by a stiff brush before the tile was fired. Fill 23019, Pit 23012, Group 694, Area N, Period 4

Other markings and oddities

One piece of tile spall, probably not a box flue tile, had two lightly impressed lines of combing made with a seven-toothed comb, forming a cross.

A corner of a flat tile from context 8591 was broken across what appeared to be the edge of a large hole, at least 50x25mm, located 100mm from one edge and 60mm from the other. The upper edge at least is original, but the lower part of the hole may be illusory. It is possibly the result of using poorly wedged clay that had not fused properly, and subsequently broke along a natural fracture line, leaving the impression of a hole. On the other hand, it might be a genuine hole, deliberately made for some unknown purpose.

The tesserae

Ceramic tesserae

In all, 9831 tesserae made from tile and pot were examined, weighing 209,549g. This represented a total surface area of 88,282cm², i.e. an area slightly less than 4.5x2m. The majority of the tesserae were made from tiles of various colours, with occasional use of suitable pieces of pottery. There were just ten pot sherds and fifteen sherds of amphora used as tesserae. This is a very minor component of the assemblage, and is unlikely to be significant, merely indicating that that the tessera makers were using any suitable material to hand. The tile used was predominantly tegula, but box flue tile and imbrex were also present. The average size of tessera from Elms Farm was 30x30mm. This is consistent in size with use in coarse tessellated pavements, rather than figurative pavements, which normally use smaller tesserae c. 10-15mm square. Some were rectangular, usually about twice the length of the 'standard' tessera; others were rather irregular, but most were roughly square.

Most of the tesserae came from rubbish deposits, or had been reused in prepared surfaces. The only tesserae surviving in situ comprised forty-one tesserae from Building 64 (context 18923). The tesserae, which were in two groups, were sketched in situ, but the deposit was vandalised by illegal detectorists prior to full recording. Unlike some sites, where caches of unused tesserae have been found (e.g. Rudston Villa; Stead 1980, 13), we can be confident that the bulk of the tesserae found had originally been used in floors or other prepared surfaces. Most of the ceramic tesserae from the site showed signs of use, in the form of surface wear and/or mortar. The wear was probably largely due to use in tessellated pavements, although in a minority of cases, it came about through accidental incorporation in surfaces such as roads. It should be borne in mind that absence of wear does not necessarily indicate that tesserae were unused, since some areas of a room would have had more traffic than others. Few of the tesserae from the group in situ in Building 64 showed signs of wear, suggesting that they came from a part of the floor not often walked on, such as the edge. Nor is absence of mortar necessarily significant; the tesserae from this group had no surviving mortar on them when examined.

The tile used was predominantly orange/red in colour (95% of the total number of pieces), with small amounts of cream/buff (4%) and grey (1%). The pottery and amphora tesserae were mostly in cream or buff fabrics, with a few grey and red sherds. Cream/buff tesserae were present in all periods, forming between 3.3% and 5.4% of the assemblage for each period. The dominance of orange/red tesserae suggests that there was little, if any, patterning present in the tessellated pavements on the site. The very small amount of grey tesserae certainly makes it unlikely that this colour was used as a pattern element. However, the cream/buff tesserae are present in sufficient numbers to suggest that they could have been used to pick out a simple pattern, such as a linear border, although it cannot be ruled out that they were just used randomly.

There was a heavy concentration of tesserae in Area J, in all periods (76% of the total number). The further away from Area J, the fewer tesserae there were, with the exception of Area R. The tesserae in the latter area were mostly connected with episodes of dumping of material that may have originated outside the excavated area. There were, however, no outstanding differences between the groups from R and those from the other areas, although cream/buff tesserae were absent from R, and none had mortar present.

Figure 501: Chronological distribution of tesserae

Deposition of tesserae began in Period 3, though on a small scale (Figure 501). Most of the early tesserae had evidence of wear, and therefore they do not derive from construction of floors. It seems most likely that, as will be suggested for the tile, they were brought onto the site from elsewhere. Over three-quarters of the tesserae came from later and latest Roman contexts. Given that the bulk of the tesserae must have derived from tessellated floors, and presumably represent demolition deposits, these periods must represent the date of the disuse/destruction of the floors, not their construction. The distribution for Periods 4 is principally round the northern edge of the temple precinct and along the northern edge of Road 4, with a scatter of finds to the north and south. All the tesserae in Area R from datable contexts are from Period 4. The distribution for the later Periods 5-6 is somewhat more compact, with fewer outliers. The distribution along the northern edge of the temple precinct and the line of Road 4 is not dissimilar to the earlier period, but there is a noticeable cluster of findspots round Building 64. It is likely that this represents debris from the pavement within this building, a small part of which was still articulated.

Fifty-nine contexts contained tesserae that retained traces of mortar or other material used as the base for the floor. In some cases, more than one layer was present, suggesting either repair, or possibly grouting. There was little discernible chronological difference in the range of mortars used; the predominant material used as a base for was a fine white mortar, sometimes with a thin layer of clayey material or sandy mortar on top. Opus signinum was present on tesserae in a few cases, but this was probably due to reuse as coarse building rubble, particularly within the precinct wall.

In conclusion, then, there was at least one building that had a plain tessellated floor made from orange/red tiles, a small part of which survived in situ. There is no evidence for any figured mosaic floors on the site.

Stone tessera

There was a single cube of off-white stone, which may have been a tessera, from context 5228. The stone type was probably limestone. The retrieval of stone artefacts from the site was generally good, and as this is the only possible stone tessera identified, it seems unlikely that they ever formed a significant component of the tessellated pavements on the site.

Glass 'tesserae'

Two opaque blue glass 'tesserae' were recovered; as with the stone tessera, it seems unlikely that they ever formed a significant component of the tessellated pavements on the site.

'Chimney pots'

The 'chimney pots' from Elms Farm were in orange-red or brownish, fairly fine textured, tile fabrics, sometimes with sanding on the edges. They were better finished inside than the tubes, and while it is possible that some of the tubes may be parts of 'chimney pots' exhibiting no distinctive features, it seems unlikely. None of them had traces of sooting. There were seven fragments, from six contexts, none of them very large. The recognisable features present are the edge of an arched cut-out, and part of a 'chimney pot' with triangular apertures. Two of the pieces have broken across small holes, 10-11mm in diameter, and set about 40mm apart. The presence of small pierced holes is unusual.

The function of this class of chimney-like object is still uncertain. Past suggestions include roof finials, ventilators and votive lamp chimneys. They were first discussed at length by Lowther (1976), who termed them 'chimney pots or finials', and distinguished two groups: group A were made from pottery fabrics, and group B from tile fabrics. More recently, the large group of 'chimney pots' from Piddington Villa, Northants., has been discussed by Ward (1999, 26-40). The use of different fabrics may be significant, with those in pottery fabrics potentially having a different function. Certainly, the use of tile fabrics for group B suggests that they were likely to be roof furniture, borne out by the existence of an example from Norton, E. Yorkshire, which incorporates an imbrex at its base. The general lack of sooting, however, suggests that they may be ventilators rather than chimney pots as such. The group A 'chimney pots', which are in finer fabrics, and which tend to be more elaborate, would seem like better candidates for votive lamps or incense burners, but again, there is little evidence for the sooting to be expected on the inside of a lamp cover - more so than on a chimney pot, since a lamp cover would be in closer proximity to a source of heat and smoke.

The Elms Farm 'chimney pots' add little to the debate. Although there are examples from other temples, they can be found on a variety of site types, and none of the Elms Farm fragments are from the immediate vicinity of the temple. They are therefore unlikely to be connected with religious ceremonies. As noted elsewhere, there is little evidence for tiled roofs on the site, and they cannot be tied in to a particular building. In fact, their distribution is very scattered, and they are not even associated with a particular area or zone of the site.

Figure 502: Chimney pots, 62-65

62. Two pieces from a ?'chimney pot', probably both the same object. The fabric is rather soft, and somewhat abraded, and the insides are well finished. The larger piece is illustrated. a) Small fragment with a straight edge. External diam. c. 110mm, T. 15mm. Wt. 26g. b) Fragment with the edge of a curved cut-out, with traces of two pierced holes, 67mm. Above the top of the arc, and placed symmetrically either side of it. The diameter of these holes seems to be about 11mm, although it is difficult to be certain, as so little of their edges survive. External diam. not measurable, Th. 20mm, Ht 105mm. Wt 140g. Fill 3558, Pit 3565, Group 889, Area W, Period 3

63. Base of a ?'chimney pot'. A fragment from a tube, with a straight rim, with a protusion on one side. This is now broken, and its original shape is unknown. It is not part of an integral ridge tile, as for example, the finial from Norton (Lowther 1976, pl. II), as the base of the protrusion is flush with the base of the tube. It could, perhaps, be part of the base of an incense burner, similar to those from Coventina's Well (Allason-Jones and McKay 1985, 41), though in this case less ornate. The outside is well smoothed, but rather irregular. Internal diam. c. 120mm, Th. 15mm, Ht 94mm. Wt 164g. SF8411, Layer 6025, Group 573, Area H, Period 5-6

64. Not illustrated. Roughly rectangular fragment, c. 60x35mm. Well finished inside and out. The outer face has striations that are probably caused by sand dragged over the surface during smoothing, rather than deliberately scratched lines. One of the short sides has a straight edge, now damaged. This may have been the base, or may have been the edge of a cut-out. It has broken across two holes, 48mm above the straight edge. One has a diameter of c. 10mm, the other is merely a trace. The holes resemble those on the fragment from 3558, but this is unlikely to be part of the same object, as the fabric is browner. The diameter is not measurable. Wt 45g. Layer 10255, Group 840, Area F, Period 5

65. Fragment from a 'chimney pot' with a triangular opening. The fragment has a central hole, and is decorated with incised lines parallel to the sides, with a row of short oblique lines adjacent. There is a scar where the flange has broken off, with a straight edge below this. It is not possible to say which way up this piece was orientated. Max. external diam. c. 100mm, Ht 55mm. Wt 59g. Cleaning layer 13316, Area I, not phased

66. Not illustrated. Fragment with two straight parallel edges, in a well-fired fabric with the inside and outside fairly well finished. The inside is slightly sandy. External diam. c. 120mm, Th. 22mm, Ht 66mm. Wt 98g. Fill 15515, Pit 15514, Group 696, Area M, Period 4

67. Not illustrated. Fragment from a ?'chimney pot' with one sanded straight edge. The edge opposite this may be original, forming the base of a cut-out; however, it is somewhat abraded. External diam. not measurable, Th. 13-19mm, Ht 83mm, Wt 64g. Layer 24298, Group 3044, Area M, Period 3

Window glass by J. Compton and S. Worrell

Ninety-six fragments of window glass were recovered from forty-seven contexts, thirteen of which were in Area R, to the north of the water channel. There are forty-nine fragments of cast blue/green window glass, ten with edges, some with tooling marks and one with mortar attached, plus twelve cast pale green fragments, five with edges and one with a corner. Twenty-seven fragments are cast greenish/colourless, six with edges, and there are five cast colourless examples. In addition, there are two blue/green fragments of blown window glass and one greenish/colourless fragment. Cast window panes are the most commonly found type, in main usage through to the 3rd century. Blown glass window panes, made as a cylinder and then cut and opened out flat, were not widely adopted until the 4th century (Cool and Price 1995, 437). Two of these examples from Elms Farm are from late 4th-century contexts, the third is from a Period 4 context in Area R. There is no window glass in any context earlier than Period 3.

Although the total quantity of window glass is small, some forty-eight of the fragments recovered come from peripheral Area R, representing over 50% of the assemblage. Nearly all occur in Period 4 contexts, perhaps indicating deposition in a single act. The window glass would appear to have been dumped from elsewhere, as there does not appear to be a building in the vicinity from which the glass could have derived. Very little (32%) of this window glass is of the common blue/green variety, most comprise greenish colourless fragments. The significance of this is not clear.

A quantity of window glass was recovered from the adjacent Crescent Road excavations (Wickenden 1986, 28), most of which came from two features in close proximity to Elms Farm Area R. The features in question are a Late Roman pit and an SFB, indicating that the activity may be contemporary with the deposition of the window glass in Area R. Roman window glass in Essex, outside Colchester, is fairly uncommon and always appears to be associated with an identified building. For instance, most of the window glass from Chelmsford came from the vicinity of the mansio (Drury 1988, 114). A large amount of window glass came from contexts associated with a stone building at Ivy Chimneys, Witham (Turner 1999, 123) and the villas at Chignall St James (Clarke 1998) and Great Holts, Boreham (Germany 2003) also each produced small quantities of window glass. It is reasonable to suggest therefore, that a substantial building existed not too far from the excavated area at Elms Farm.

Building Stone

Architectural stone

Three pieces from the site may be carved architectural stone (as opposed to coarse building rubble, which is essentially unworked). Identifications are by Dr G.K. Lott of the British Geological Survey (GKL).

Figure 503: Building stone, 68-70

68. Joining fragments from three contexts in the same pit, forming a slab with shaped edges. One face is smooth, the other broken. The shape approximates a right-angled triangle, with the hypotenuse forming a concave curve. The top face is chamfered on one edge, and there is a groove across the face, from this edge, running below the curved edge. Possibly architectural. Max. surviving Th. 67mm. Identification (GKL): Lithic sandstone, Upper Carboniferous (Millstone Grit Group or Coal Measures). Wt 1537g. Fills 5832/5841/5864, Pit 3805, Group 444, Area J, Period 6

69. A slab fragment, with the edge probably reused as a sharpening stone. The shape is possibly a segment of a circle, and it is possible that this was a decorative stone, possibly used in a building (probably not a quern). One face is very smooth, the other is smooth in patches. Th. 25-31mm. Identification (GKL): Quartzose sandstone, possibly Millstone Grit Group or Coal Measures. Wt 350g. Layer 16187, Group 573, Area H, Period 5-6

70. Millstone Grit. A quarter circle, with the curved edge well finished, the other edges broken. One face is well finished. The diameter is far too small for this to be the original quern edge, but it could be part of a door pivot stone. Max. Th. 45mm. Wt. 266g. Machining layer 11000

Unworked building stone

Much of the 'unworked' stone collected from the site had been used as building rubble, and some may have been very crudely shaped. The stone collected was a sample only, and analysis by quantity is therefore unreliable. However, it is possible to make some general comments on spatial and chronological distribution.

Stone used as building rubble includes stone types that are not local to the site, in particular, Kentish greensand and septaria, the latter being the commonest building stone on the site (Table 98). Septaria is found in deposits along the north Essex and Suffolk coast, and was used extensively in Roman Essex. The greensand may have been arrived on site primarily as ballast in trading vessels, rather than being deliberately imported as a building stone; the tufa is probably also from Kent. Utilisation of stone for coarse building material would have been opportunistic, given the lack of good building stone in Essex.

With both septaria and greensand, it is difficult to tell whether it has been deliberately shaped: in the case of the septaria, because it tends to break naturally into irregular blocks, and in the case of the greensand, because the surface has usually eroded. Some of the tufa certainly seems to have been cut into rough blocks, and it is likely that this was true of the greensand as well. Septaria does not cut well, and was probably never neatly trimmed.

A possible alternative use for septaria is noted in passing. Morgan (1992a) suggests that septaria could have been used as a source for hydraulic lime manufacture; analysis of samples of septaria from Colchester showed a lime content of about 48%. It is impossible to determine whether septaria was used in this way at Elms Farm.

Other stone used as building rubble includes flint nodules, various sandstones (probably mostly derived from local erratics), tufa, shelly sandstone, quartzite and sarsen. Table 99 lists the number of contexts from which unworked stone of each type was collected; the list may include some non-building stone, such as the burnt flint nodules.

Table 98: Types of unworked stone collected at Elms Farm
Type of stone No. of contexts
Septaria 216
Greensand 85
Burnt flint nodules 74
Sandstone 71
Natural pebbles 42
Chalk 28
Not identified 23
Flint nodules 19
Quartzite 14
Tufa 12
Sarsen 11
Limestone 11
Grit 3
Ferruginous sandstone 2
Ferruginous conglomerate 2
Kentish Ragstone 1
Pumice 1

Table 99 lists the approximate number of contexts containing unworked stone of all types within each area (this includes burnt flint and burnt pebbles, which are unlikely to be building stone). It is clear that the majority of the unworked stone came from the vicinity of the temple (Area J), with very little coming from the outlying parts of the site, Areas Q and R in particular.

Table 99: Unworked stone; approximate number of contexts per area
Area No of contexts
D 12
E 7
F 26
G 20
H 59
J 172
K 12
L 15
M 26
N 19
P 6
Q 1
R 4

It was assumed in the assessment that much of this material had been used as coarse building rubble, and further postulated that different stone types may have been used at different periods. When the proportions of contexts by phase containing septaria and greensand are compared, however, the evidence suggests that, for these two types of stone at least, the deposition rates were remarkably similar (Table 100).

Table 100: Septaria and greensand: comparison of numbers of contexts and percentages of contexts by period (discrete periods only)
  Septaria Unworked Greensand
Period No. % No. %
2 7 4 5 8
3 35 21 12 20
4 25 15 13 22
5 23 14 7 12
5-6 22 13 9 15
6 44 27 14 23
7 7 4 0 0
Total 163   60 -

Both types of stone occur in small quantities in Period 2 contexts, and, apart from Period 7, show a fairly consistent pattern of deposition. For both, there is a drop-off in deposition in Period 5, though this is only slight for the septaria. All the Period 7 septaria comes from the post-medieval fence-line in Areas H and J, and probably represents stone robbed from the precinct wall and used as packing in some of the post-holes.

The presence of septaria in seven Period 2 contexts initially suggested that it was already being brought onto the site before the Roman conquest. However, four of the contexts with septaria contain at least some Roman material. One Period 2A context (post-hole 18734 in Area J) contained a substantial amount of septaria, over 78kg, but it is probable that the context number was a mistake for 18739, which was a Period 4 wall. The presence of greensand prior to the conquest is not unexpected, as there was an established trade in greensand querns into this part of the country from the Bronze Age onwards. Septaria, however, is a soft and brittle stone only suitable for use as coarse rubble, and its presence in such quantity in a Late Iron Age context could be considered unusual.

Fasteners and fittings


The following are copper alloy unless otherwise specified

Bell-shaped studs

Bell-shaped studs are common finds on Roman sites, and can be divided into two groups, those with integral, rectangular, shanks, pierced at the end, and those with inserted iron shanks, or with integral round or square-sectioned shanks. The type is discussed by Allason-Jones (Allason-Jones and McKay 1985, 30) with regard to the examples from Coventina's Well, citing examples from Corbridge in contexts dating before AD 90, and a large group from Piercebridge, a late 3rd-4th century fort. This type of stud therefore appears to be an early introduction into Britain, and continues in use throughout the Roman period. They are probably primarily box fittings, though some have shanks which seem too long for this purpose. Those with integral rectangular shanks are probably lock pins.

There were thirteen bell-shaped studs with circular or square-sectioned shanks from Elms Farm, and three with rectangular-sectioned shanks (described with the locks and keys). Their manufacture on the site is attested to by the presence of two mis-cast studs (function category 15, nos 13 and 14). Several of the others appear to be rather poorly finished. The heights of the studs (without the shank) are 5.5-28.5mm, with an average of 15.7mm, and the diameters are 14-44mm, with an average of 23.8mm. Five examples have inserted iron shanks with a square section, one has an inserted iron shank with a circular section, and on two the shank is missing. Five studs have integral copper-alloy shanks with square sections.

Only one is from a stratified context (no. 3, below). Three are illustrated, to show the variety of size.

Figure 504: Fastenings and fittings, 1-87

1. Bell-shaped stud. The shank is missing, but the corrosion on the back suggests a square iron shaft. There is an incised line round the back of the stud, and probably round the base. The shape is as Allason-Jones and McKay 1985, 32, no. 84. In fairly good condition, surface partly obscured by earth. Ht 15mm, diam. 17mm. SF1068, Machining layer 4000, Area A, unstratified

2. Bell-shaped stud, with a line round the base of the bell. The iron shank probably has a square section. Margaret Brooks notes that this appears to have been burnt. It comes from the pit in the centre of the putative 'lead-working' building (Building 63), and there is other burnt metal from the pit. Although it includes melted lead, there is also burnt copper alloy, and iron. Diam. 27mm, Ht 19mm. SF2070, Fill 5399, Pit 5400, 454, Area J, Period 6

3. Large bell-shaped stud with concentric moulding round the edge, shank missing. The back is not very well finished. An unusually large example in good condition. Diam. 44mm, Ht 28.5mm. SF3871, Machining layer 11000, Area A, unstratified

4. Not illustrated. Bell-shaped stud with the remains of a circular-sectioned iron shank. There is a horizontal hole through the base of the stud on one side only, possibly due to a poor casting. The surface is in poor condition and the edge damaged. Ht 16mm, Diam. 19mm. SF9526, Context 3999, Spoil-heap

5. Not illustrated. A small bell-shaped stud. The base is a plain cylinder with 4mm of a square-sectioned iron shank surviving. The top and edges are damaged. In fair condition. Stud ht 14mm, diam. 16mm. SF983, Machining layer 4000, Area A, unstratified

6. Not illustrated. Bell-shaped stud, with a square-sectioned copper-alloy shaft, broken. The back is poorly finished, with a circumferential line round part of the base. This is a relatively large example of the type. Diam. 34mm, Ht 17mm, total L. including shank 23mm. SF478, Cleaning layer 6000, Area H, Not dated

7. Not illustrated. Bell-shaped stud. A small example with a broken integral square-sectioned shank. The bell is shallow, with a very short base. The condition is fairly poor, with damaged edges. Diam. 14mm, Ht of bell 5.5mm, overall L. 13mm. SF2218, Cleaning layer 8000, Area E, not dated.

8. Bell-shaped stud, iron shank missing. The shank was probably square. Surface in fairly poor condition. Diam. 21mm, ht 16mm. SF1566, Machining layer 11000, Area A, ynstratified

9. Not illustrated. Bell-shaped stud, with a broad and rather shallow 'bell' in a thinner metal than normal. There is an incised line round the junction of the bell and base. The (iron?) shank is missing. In fairly good condition, ancient damage to edge. Diam. 27mm, Ht 12mm. SF5309, Machining layer 11000, Area A, unstratified

10. Not illustrated. Bell-shaped stud, with the stub of a square-sectioned iron shank. It has a moulded line round part of the outer edge of the bell, and a moulded groove round the base. Diam. 15.5mm, Ht 14.5mm. SF5413, Machining layer 11000, Area A, unstratified

11. Not illustrated. Bell-shaped stud with an integral square-sectioned shank, poorly cast, and possibly not finished off. The thickness of the metal is variable, and the underside rather irregular. The base is waisted. Diam. 25mm, Ht 19mm. SF3165, Machining layer 12000, Area R, unstratified

12. Not illustrated. Bell-shaped stud, with an integral, slightly tapering, square-sectioned shank which survives to a greater length than most from the site. The cylindrical base probably had a line round the middle. In very poor condition. Diam. 27mm, L. 49mm, L. of shank 35mm. SF6364, Machining layer 12000, Area R, unstratified

13. Not illustrated. Bell-shaped stud with an integral, broken, square-sectioned shank. The cylindrical base has a line round it, and the central projection is relatively tall. In fair condition, but with the surface covered in ferruginous corrosion. Diam. 22.5mm, Th. of head 14mm, overall L. 26.5mm. SF6762, Machining layer 17000, Area Q, unstratified

Flat round head

This type of stud, most of which had thin, disc heads, fared particularly badly during burial and excavation, and the majority of the examples are damaged.

None of the plain studs are illustrated.

14. Stud, circular, with nicks round the edge and a long shank bent at right-angles (Figure 504). The end of the shank has been crimped. In fair condition, with damage to the edge. This is possibly part of a button-and-loop fastener. Diam. 24mm, L. of shank 26mm. SF346, Machining layer 5000, Area J

Not illustrated

15. Sheet disc, probably a stud, with iron corrosion on the back. Diam. 30mm. SF9527, Context 3999, Spoil-heap

16. Stud. Plain circular, thin metal. Slightly domed head with a broken shank. Good condition. Diam. 27mm, Ht 9mm. SF675, Machining layer 4000, Area A, unstratified

17. Stud, ?circular, ?plain, edges damaged. Shank broken, head bent. Surface obscured by earth, fairly good condition. Max. W. (flat) 26mm, Ht 11mm. SF676, Machining layer 4000, Area A, unstratified

18. Stud, probably circular originally. In poor condition, all edges damaged. Shank broken and flattened against the back. Max. surviving diam. 12mm. SF685, Machining layer 4000, Area A, unstratified

19. Iron stud or nail with a small, round head and a round-sectioned shank, piercing a broken, slightly convex, iron sheet. On the back of the sheet are the remains of a copper-alloy object, probably a circular washer. Sheet c. 18x16mm, nail head diam. 9mm, 'washer' diam. 13mm. SF4186, Fill 4827, Post-hole 4970, Group 6006, Area K, Period 6

20. Stud, edges damaged; probably flat and circular originally. The end of the shank is missing. In fair condition. Diam. 25mm, L. 17mm. SF3195, Cleaning layer 5702, Area J, not phased

21. Stud fragment; a flat disc with a line round the edge, shank broken off. In good condition, possibly post-Roman. Diam. 20mm. SF1675, Cleaning layer 6000, Area H, not phased

22. Stud, with large circular head, shank flattened against the back. Plain, with slight damage to the edge, in fairly good condition. Diam. 33mm, shank L. 17mm. SF2733, Layer 6269, Group 581, Area H, Period 5-6

23. Iron stud, with a large flat circular head, and a short shank, possibly complete. Head diam. 30mm, L. 20mm. SF2363, Layer 6350, Group 2002, Area H, Period 2B

24. Stud fragments in poor condition and somewhat battered; all the edges are damaged, but it probably had a flat, circular head. Max. surviving diam. 18mm, L. 17mm. SF1465, Fill 10000, Pit 10062, Group 811, Area E, Period 4

25. Stud, almost flat, circular head. In fair condition, with some damage to the edge, and the point of the shank missing. Diam. 31mm, L. 23.5mm. SF2524, Machining layer 11000, Area A, unstratified

26. Stud, flat head, probably originally circular, but all edges and the shank are broken. In fair condition. Surviving diam. 29mm, surviving Ht 6mm. SF4373, Machining layer 11000, Area A, unstratified

27. Stud. An oval disc, with no definite original edges, with the stub of a shank on the back. Head 12x10mm. SF5080, Machining layer 11000, Area A, unstratified

28. Stud with small, thick, round head and a short, thick shank, complete as buried. In good condition. Head diam. 10mm, Th. 3mm; Ht 7mm, shank diam. 6mm. SF5321, Machining layer 11000, Area A, unstratified

29. Stud, with a circular head, bent, and damaged round the edge, shank incomplete. The surface bears what appear to be a number of dots, but these are just excrescences owing to corrosion. The underside has a low, moulded ridge running round it, 2mm in from the edge. Diam. 34mm, Ht 20mm. SF5683, Machining layer 11000, Area A, unstratified

30. Stud, with ?round flat head, most of the edge broken, and shank incomplete. There is a slight ridge round the edge underneath. The shank is slightly bent. Original diam. c. 31mm, Ht 15mm. SF5693, Machining layer 11000, Area A, unstratified

31. Stud, flat head, edges damaged, surface obscured. Incomplete integral shank. Max. surviving diam. 15mm, ht 5mm. Fill 13639, Pit 13640, Group 594, Area I, Period 3A

32. Stud, with a flat head, all edges broken, and incomplete shank. Max. surviving diam. 12mm, ht 6mm. SF6449, Fill 13798, Pit 13771, Area I, Period 3A

33. Stud with flat disc head, slightly down-turned round the edge. Shank incomplete. In poor condition. Diam. 14mm, ht 6mm. SF7364, Fill 13825, Pit 13771, Group 594, Area I, Period 3A

34. Part of a small stud or rivet with a flat head. The edges are damaged, and the shank mostly missing. Head is now 6x5mm. SF8376, Fill 15288, Pit 15289, Group 239, Area M, Period 2B

35. Stud, circular, with the broken shank set off centre. In poor condition, with little surface left. Diam. 14mm, Ht 9mm. SF6814, Machining layer 17000, Area Q, unstratified

36. Stud, with a sheet head, probably original circular, but now broken all round the edge. The shank is bent at right angles, and the point is missing. Surviving L. 16mm, surviving diam. 10mm. SF7777, Prepared surface 21591, Group 402, Area J, Period 3B

37. Stud, incomplete sheet head, probably slightly domed. Diam. c. 24mm, L. 10mm. SF7807, Cleaning layer 24058, Area M, not phased

Rectangular head

38. Not illustrated. Iron stud, with a flat rectangular head with rounded corners, and short shank, probably broken. 20x24mm, shank L. 12mm. SF8045, Spread 21568, Group 378, Area J, Period 3B

Domed head

39. Stud. Large mushroom head with a dished centre (Figure 504). The shank is bent. In fair condition. Diam. 36.5mm. SF1946, Cleaning layer 4757, Area K, not dated.

40. Copper-alloy stud, with a solid mushroom-shaped head with a reel below. The broken iron shank, probably circular section, is set into a rectangular hole in the base. The object is not very well modelled, and more oval than round. 12x13mm, ht 12mm. SF5424, Machining layer 11000

Not illustrated

41. Stud, most of edge missing. Domed sheet with a rolled edge. The end of the shank is turned over. Diam. 18mm, shank L. c. 10mm. SF5623, Layer 6790, Group 484, Area H, Period 3

42. Sheet fragment, possibly originally a disc, but all edges now broken. There is a domed stud through the middle, shank broken. Probably a box fitting. In poor condition, with little surviving surface, and extensive mineralised organic material on the top. Diam. c. 19mm, stud diam. 8mm. SF3517, Fill 13048, Pit 13158, Group 641, Area I, Period 4

43. Domed stud head, with the stub of the shank on the back, broken off in antiquity. Fair condition, recent damage to edges. Diam. 22mm, Ht 7mm. SF2235, Spread 10259, Group 818, Area F, Period 4-5

44. Boss or stud; slightly domed copper-alloy disc with lead on the back, and possibly an iron shaft. In fair condition, distorted by the corrosion of the lead. Diam. 24mm, ht of boss, 6mm, total Th. 13mm. SF1885, Fill 5210, Pit 5209, Group 442, Area J, Period 6

45. Stud, probably in a high lead alloy. It has a solid plano-convex head, probably with moulded decoration on top. Most of the surface has flaked off. The shank is probably incomplete. L. 12mm, Diam. 10mm. SF9536, Context 3999, Spoil-heap

46. Stud. Probably plain, circular, and slightly domed, though the edges are now damaged, and the surface obscured. The shank was broken in antiquity. Max. surviving diam. 25mm. SF1062, Machining layer 4000, Area A, unstratified

47. Stud head, made from sheet; hollow mushroom-shaped, probably post-Roman. Diam. 30mm. SF7201, Machining layer 4000, Area A, unstratified

48. Stud, most of shank missing. It has a circular, slightly domed, head, slightly flattened on one edge in antiquity. The back has a very low concentric ridge halfway across. In good condition. Diam. 33mm. SF5138, Cleaning layer 5602, Area I, not phased

49. Stud, with an integral rivet with a ?circular foot. The stud is slightly domed, with all edges broken. Probably circular originally. In fairly poor condition. Ht 8mm. SF1637, Cleaning layer 6048, Area H, not phased

50. Domed stud, shank broken. In poor condition, edges broken. Surviving diam. 12mm, ht 9mm. SF5810, Cleaning layer 10905, Area N, not phased

51. Stud with a slightly convex head, probably originally circular. All edges are broken, and the shank is incomplete. Max. surviving diam. 24mm, L. 14.5mm. SF3023, Machining layer 11000, Area A, unstratified

52. Stud, mushroom-headed, with a short, round-sectioned shank. In fairly poor condition, with slight damage to the edge. The surface has a layered effect on part of the edge, possibly the result of a poor casting. Diam. 15mm, Ht 7mm. SF3662, Machining layer 11000, Area A, unstratified

53. Iron. Probable mushroom-shaped stud head. It has traces of copper alloy on the surface, probably part of another object. Diam. 25mm, Ht 15mm. SF5360, Machining layer 11000, Area A, unstratified

54. Stud or rivet with a solid mushroom head and circular sectioned shank. There was probably an integral rove, now mostly missing. In poor condition, edges damaged. Diam. 16mm, Ht 13mm. SF6367, Machining layer 1200, Area R, unstratified

55. Plano-convex stud, with two broken shanks on the back, possibly a broken loop. The metal is very dark, and in good condition. This is possibly post-Roman. Diam. 10mm, Th. 3mm. SF6496, Machining layer 1200, Area R, unstratified

56. Convex sheet fragment, possibly with the stub of a shank on the back. Probably a mushroom-shaped stud head. 9x8mm. SF7570, Cleaning layer 21500, Area J, not phased

Concave top

57. Not illustrated. Slightly concave round head. It has a short shank, with the point damaged. In fair condition, with slight damage to the edges, and the surface obscured by earth. Diam. 32mm, L. 8mm. SF3574, Machining layer 11000, Area A, unstratified

58. Not illustrated. Round, slightly concave head, with a low concentric moulding just inside the edge on the underside. The shank is damaged. In good condition. Diam. 30mm, L. 15mm. SF3575, Machining layer 11000, Area A, unstratified

59. Not illustrated. Incomplete sheet head, probably a dished disc with rolled edges. In fair condition, whole of edge damaged, shank incomplete. Diam. 17mm, Ht 8.5mm. Original diam. c. 20mm. The type is not very common, though there are parallels from Segontium (Allason-Jones 1993, 178, no. 126), dated C1 to late C3, and Vindonissa (Unz and Deschler-Erb 1997, Taf. 73, no. 2204). SF5338, Machining layer 11000, Area A, unstratified

60. Not illustrated. A rather thick inverted conical head, dished in the centre. The broken shank has an ?oval section. In poor condition, with all edges broken in antiquity. Max. surviving W. 23mm, Ht 19mm. SF5612, Machining layer 1200, Area R, unstratified

61. Stud with a cupped head and a round-sectioned shank, the point missing (Figure 504). In fairly good condition. There is no sign that the cup ever contained enamel or a stone. There is an almost identical stud from Colchester, also with no inset surviving, from a pre-Boudiccan context (Crummy 1992c, 229, no. 301). The Colchester stud differs in that it has a square-sectioned shank. Diam. 18mm, L. 20mm. SF7059, Machining layer 17000, Area Q

Globular head (Not illustrated)

62. Globular head, probably from a furniture stud, since the shaft appears to be tapering sharply, but possibly from a hair-pin. In poor condition. Diam. 7mm, surviving L. 11mm. SF5378, Machining layer 11000, Area A, unstratified

63. Rod with a sub-globular head, other end broken. This is probably a decorative stud rather than a hairpin, although it would be more normal for a stud to have a square-sectioned shank. L. 29mm, head diam. 13mm. SF3054, Machining layer 11000, Area A, unstratified

64. Stud head? A sub-spherical head with a broken shank, probably from a furniture stud rather than a hairpin. Head diam. 7.5mm, L. 14mm. SF5571, Machining layer 11000, Area A, unstratified

Other studs

65. A flat, circular sheet with a rectangular projection on one side, and a curved and broken shank (Figure 504). This is apparently its original shape, although the edge is flaked all the way round, and it is not possible to be completely certain. 13.5x15mm, Ht 6mm. SF677, Machining layer 4000

66. Bar stud, with an integral central rivet. The top has three moulded elements, with a slightly hollow back at one end. Possibly military. L. 22mm, W. 9.5mm. SF5709, Machining layer 11000

67. Peltate stud with two integral rivets. The back is hollow. In fair condition, with damage to one edge. L. 22mm, W. 23mm. Possibly military. SF5719, Machining layer 11000

68. Stud or rivet, in very grey metal. It has an oval head with a line round the edge and a central dimple and boss. The shank is a robust, solid cylinder, relatively long. The surface is flaking. Head 21x19mm, L. 17.5mm. SF2486, Machining layer 4000, Area A

69. Stud, with a symmetrical complex curvilinear shape, one end incomplete. There are two integral rivets on the back. Possibly military. L. 38mm, W. 13mm. SF3169, Machining layer 12000, Area R

70. Stud. Cast, rectangular in plan, and a segment of a circle in side view. The underside is hollow. The curved top has a central crest with a rib either side. It has two spikes integral with the top, set in the middle of each long edge. These are triangular, and project slightly at the top. In fair condition, with modern damage to the top. 16x12mm, Ht 16mm. SF8151, 3999, Spoil-heap

71. Not illustrated. Pointed rod fragment with a square section. Stud or nail shank? No surface surviving. L. 20mm. SF2230, Layer 10248, Group 840, Area F, Period 5

72. Not illustrated. Stud. Conical, all edges broken, shank broken. In poor condition. Max. surviving diam. 12mm, ht 6mm. SF5918, Cleaning layer 15024, Area M, not phased

73. Not illustrated. Stud, with a square, flat head and a thick shank without a point. L. 17mm, head W. 12mm, Th. 4mm. SF9527, Context 3999, Spoil-heap

Copper-alloy bosses

Following the terminology used for the Colchester finds reports, the term boss has been used only for hollow convex objects with no integral shaft. Domed sheet bosses filled with lead solder, such as nos 74-76, below, were certainly used as box fittings (for example, on a box from the Butt Road cemetery, Colchester; Crummy 1983, 85-7), though other decorative uses cannot be ruled out.

74. Not illustrated. Domed boss, possibly with white-metal coating. In fairly good condition, edge damaged. The back has the remains of ?lead-based solder, and possibly bone, although the latter material is not necessarily associated with the object. Diam. 19mm, Ht 6mm. SF1179, Context 4364, Hearth construction 4378, Group 742, Area K, Period 5

75. Not illustrated. Domed boss with ?lead-based solder on the back. In fair condition, edge damaged. Diam. 18mm, Ht 7mm. SF1648, Cleaning layer 6000, Area H, not phased

76. Not illustrated. Domed boss, the back filled with a lead-based solder. The lead corrosion has cracked and distorted the boss. Diam. 20mm, Ht 5.5mm. SF5778, Fill 10891, Pit 10910, Group 676, Area N, Period 5

77. Hollow, cast, hemispherical boss with a slight flange round the bottom, and a hole in the top. It has a square-sectioned iron shank through the hole, presumably originally with a head, which is now missing. Diam. 25mm, Ht 17mm. SF2963, Machining layer 11000

Figure 504: Fastenings and fittings, 1-87

Copper-alloy knobs

Figure 505: Fastenings and fittings, 88-105

78. Moulded knob, with an oval-sectioned iron shank set into what appears to be a short integral tube in the middle of the base, now only partly surviving. 4000, SF588 is almost identical, apart from the tube on the base. Surface powdery. Ht 26mm, max. Diam. 17mm. SF304, Machining layer 4000

79. Not illustrated. Knob, with an iron shank broken flush with the base. Almost identical to 4000, SF304. In good condition. Ht 25mm, max. Diam. 17mm. SF588, Machining layer 4000

80. Campanulate knob with mouldings on the shank, which appears complete. There is a shallow hole in the bottom, and the base has concentric grooves. There is a strip of a different alloy, in poor condition, wrapped incompletely round the base of the head. The shape is unusual, and the moulding on the shank suggests that this is not an ordinary knob. XRF analysis showed that the knob was a quaternary alloy (Copper/tin/lead/zinc), while the band was a leaded bronze (copper/tin/lead). L. 34mm, max. Diam. 24mm. In good condition, with a blackish patina. SF668, Machining layer 4000

81. Spherical, with a button on the top and a cylinder below. Very little of the shaft survives. Ht 16mm, max. Diam. 12.5mm. SF917, Machining layer 4000

82. Not illustrated. Spherical knob, with an uneven surface, and traces of a square-sectioned iron shaft. The knob is relatively heavy for its size, probably indicating a high-lead alloy, and in very good condition. Diam. 16mm. SF1866, Machining layer 4000

83. Not illustrated. Object in leaded copper alloy, in poor condition. The corrosion of the lead content has caused the object to become misshapen. It was probably a boss or terminal knob with a circular section, and a domed or conical top with a dimple in the middle. Diam. c. 21mm, Ht 16mm. SF6506, Machining layer 11000

84. Mushroom-shaped knob with a circular moulding on the top, the centre dished. The broken iron shaft has a rectangular section. The hollow underside is poorly finished, with the mould marks left on. In good condition. Overall L. 38mm, L. of knob 23mm, diam. 31-33mm, shaft section 10x5mm. SF5148, Cleaning layer 5603, Area I, not dated

85. Campanulate knob with a broken, square-sectioned iron shank. In good condition, but with iron corrosion obscuring part of the surface. The iron may be the remains of a wire wrapped round the neck of the knob. Ht 30mm, max. Diam. 19mm. SF5389, Machining layer 11000

86. Knob, probably a box fitting. It has a moulded body with a recessed top with a central hole, with a slight collar round it. The hole evidently completely perforates the object, as the iron shank is visible in the top. The other end of the square-sectioned shank is broken. Ht 16.5mm, Diam. 17.5mm. SF5390, Machining layer 11000

87. Knob with lenticular section, in a heavily leaded bronze. The top has eight moulded petals, poorly spaced round the object and of varying height There was a separate shank in a different alloy, completely perforating the knob, and visible in the centre of the top, but not in the base. The tubular base of the knob contains lead, or a high-lead solder, fixing the shank in place. There is a rather crudely executed circumferential groove round the base. Diam. 27mm, Ht 16mm. SF5711, Machining layer 11000

88. Campanulate terminal with the stub of a ?square shank. The top of the knob is missing. The surface is fair where it survives, but there is some damage to the surface. Ht 23mm, max. Diam. 6mm. SF5910, Fill 15046, Ditch 15045, Group 410, Area M, Period 3

Copper-alloy mounts

89. Pelta-shaped mount with bent back loop. The loop is unusual; one would normally expect an integral rivet. The bend does not appear to be original. In good condition, slightly damaged. W. 26mm, L. 16mm. SF1065, Machining layer 4000

90. Mount, complex shape, with two long integral rivets on the back bent into rectangular loop. The surface is in poor condition. L. 26mm, W. 16mm. SF1842, Machining layer 4000

91. Not illustrated. Small peltate mount, distorted, with two integral rivets on the back. Similar to peltate stud SF5719 (no. 67, above), though smaller. Possibly military. W. 19mm, L. 20mm. SF2500, Machining layer 4000

92. Circular mount with concentric moulding round the centre and edge. The central hole contains the remains of an iron pin, broken off flush with each side of the disc. In fair condition. Diam. 38mm. SF5146, Cleaning layer 5603, Area I, not phased

93. Oval mount in the shape of a moustachioed face with curly hair and beard. The hollow back is filled with lead. Possibly a bowl mount. The edge is damaged and cracked. Ht 29mm, W. 26mm, Th. 16mm. SF4285, Machining layer 11000

94. Cast disc, with two concentric cells on top, and a central hole. It was possibly originally enamelled, though no trace of this survives. The back of the object seems to have been dented in antiquity, causing a crack on the upper surface. Patches of hard greyish matter adhering to the back are probably a build-up of corrosion between the roundel and whatever it was attached to. Diam. 30mm, hole diam. 4mm. SF5407, Machining layer 11000

95. Not illustrated. Mount fragment, gilded copper-alloy sheet. One bent integral rivet survives. Only one end is complete, and the original shape is uncertain. The complete end is rectangular, with a ?lenticular adjacent element. This is possibly medieval rather than Roman; gilding was relatively common on medieval bar mounts, although the integral rivet is not such a common feature. L 21mm, W. 9mm. SF6771, Machining layer 17000, Area Q

Figure 505: Fastenings and fittings, 88-105

Copper-alloy tacks

The term tack has been used for objects with small heads of various shapes (mushroom-shaped, globular or disc-shaped) and short, pointed shafts. Those with longer shafts have been classified as nails.

96. Not illustrated. Hollow mushroom-shaped head. Possibly post-Roman. In fair condition, distorted. Head diam. 10mm, L. 18mm. SF8194, Context 3999, Spoil-heap

97. Not illustrated. Hollow mushroom-shaped head. Possibly post-Roman. In good condition, brown patina. Head diam. 10mm, L. 18mm. SF8195, Context 3999, Spoil-heap

98. Not illustrated. Hollow mushroom-shaped head. Possibly post-Roman. Fair condition, point missing. Head diam. 14mm, L. 7mm. SF8196, Context 3999, Spoil-heap

99. Not illustrated. Five tacks with globular heads. a) Incomplete. L. 14mm, Head diam. 7.5mm. b) Complete. L. 15mm, Head diam. 6mm. c) Incomplete. L. 17mm, Head diam. 8mm. d) Incomplete. L. 15mm, Head diam. 9mm. e) Complete. L. 23mm, Head diam. 8mm. SF9521, Context 3999, Spoil-heap

100. Not illustrated. Ball-headed tack with a chisel point. In good condition. L. 20mm, head diam. 8mm. SF1026, Machining layer 4000.

101. Not illustrated. Tack, with a small, flat disc head, possibly incomplete. In poor condition. L. 11mm, Head diam. 4mm. SF1991, Unknown context 5385, Area J, not phased

102. Not illustrated. Copper-alloy tack, corroded onto an iron strip but not necessarily associated with it. The length is complete, but half of the flat, round head missing. L. 11mm, Head diam. 3.5mm. SF5296, Cleaning layer 5543, Area J, Period 5-6

103. Not illustrated. Ball-headed tack, very poor condition, no original surface left. Probably complete. L. 11mm, Head diam. 4.5mm. SF6307, Fill 8802, Pit 8801, Group 678, Area P, Period 6

104. Not illustrated. ?Tack in very poor condition. Slightly expanded head. No original surface left. L. 14mm, Head diam. 5mm. SF3277, Fill 9407, Gully 9598, Group 804, Period 4

105. Tack with a sub-globular head and concave top. The shank is slightly bent, and the end is missing. Although cupped tops often held glass insets, the concavity on the top of this example seems too shallow to have held an inset. Probably a box fitting, or from furniture upholstery. The surface is fairly poor. Head diam. 10mm, Th. 7mm. Overall L. 20mm. SF1478, Fill 10017, Pit 10067, Group 837, Area E, Period 5

106. Not illustrated. Tack head, spherical, shank broken, surface flaking. L. 13mm, Diam. 9mm. SF6721, Machining layer 11000

107. Not illustrated. Domed head, shank missing. Diam. 10mm. SF7195, Machining layer 11000

108. Not illustrated. Spherical, truncated at the base, with most of the shank missing. Head diam. 10mm, L. 13mm. SF7072, Machining layer 17000, Area Q

109. Not illustrated. Probable sub-globular tack head, in poor condition. L. 12mm, Head diam. 6mm. SF7160, Machining layer 17150, Area A

110. Not illustrated. Tack with a solid hemispherical head. Complete, but in poor condition. L. 14mm, Head diam. 8mm. SF6012, Machining layer 17242, Area C

111. Not illustrated. Tack with a flattened sub-globular head. Surface in poor condition. L. 16mm, Head diam. 7mm, Th. 5mm. SF7530, Fill 20752, Ditch 20751, Group 466, Area L, Period 6

112. Not illustrated. Tack with a globular head, point missing. L. 20mm, Head diam. 5.5mm. SF7670, Cleaning layer 21615, Area J, Period 3B


Copper-alloy nails (not illustrated)

113. Small disc head, point turned up. In poor condition. L. 31mm, Head diam. 4mm. SF5625, Layer 6467, Group 3003, Area H, Period 3

114. Bent, with a small circular head. Probably almost complete. In poor condition, surface missing. L. 25mm, Head diam. 4mm. SF7363, Fill 13825, Pit 13771, Group 594, Area I, Period 3A

115. Circular head. The square-sectioned shank is bent, and the head is slightly expanded, and possibly incomplete. In poor condition, surface powdery. L. 38mm, W. of head 5mm. SF3336, Layer 5883, Group 369, Area I, Period 3B

116. Point missing and little of the head surviving. L. 26mm. SF3997, Fill 10361, Ditch 10657, Group 838, Area F, Period 5-6

117. Head damaged. Bent, square-sectioned shank. L. 22mm. SF7060, Machining layer 17000, Area Q, not phased

Lead nails by Ros Tyrrell

The function of lead nails is uncertain since such a soft metal would rule out their use as conventional nails. Two examples were found, both from the topsoil and possibly post-Roman.

118. Not illustrated. A flat, round-headed, lead nail, with a round-sectioned shaft. L. 27mm. SF5492, Machining layer 11000.

119. Not illustrated. A flat, rectangular-headed, lead nail, with a square-sectioned shaft. L. 52mm. SF5492, Machining layer 11000.

Iron nails

In total, 6098 iron nails and nail fragments were recovered from the site. Most were covered in corrosion products or concretions, and measurements were mostly taken from X-rays. It was originally intended to X-ray all the nails, but after consultation with N. de Silva and the former Ancient Monuments Laboratory (now within English heritage Intervention and Analysis), it was decided that this was not feasible. The bulk nails were therefore catalogued at the assessment stage, and a selection sent for X-ray. The following criteria were used in the selection process;

  1. All definite objects other than nails were selected.
  2. Items where there was no detail visible due to concretion
  3. Definite nails with possible unusual features
  4. All 'type D' nails, which are scarce, but very similar in shape to simple chisels.

Most of the 'nails' selected fell into the third category. Most of the small-find nails were X-rayed with the other iron small finds.

The nails were catalogued on pro formas using the standard Essex County Council type series to record shape and dimensions. Lengths were recorded for complete nails only; head dimensions of the heads were recorded for all nails. The data were input into an Excel spreadsheet.

A sample of one hundred complete nails from Area K (context within the 4000s) was analysed for the assessment. Within this group (as with most Roman sites) the most common form was that with a round, flat head (ECC type A), forming 60% of the sample, followed by nails with oval heads (19%). Hobnails were rare (2%), as they were for the whole bulk and small-find nail assemblage. However, when the nails from the soil samples were examined, it was found that they contained a higher proportion of hobnails, and a much greater proportion of the sampled contexts contained hobnails. Out of the thousand or so contexts with bulk or small-finded nails, only fifty contained hobnails, whereas eighty-six of the 192 sampled contexts containing nails produced hobnails. The type of context selected for bulk samples (e.g. rubbish pits) may be more likely to contain hobnails, but it is more probable that, in general, there was poor recovery of hobnails using standard digging techniques. This is no doubt partly due to their similarity to the small lumps of iron pan that occurred in many contexts, which themselves were often mistakenly collected as iron objects.

Figure 506: Nail forms

Twenty-five different forms were identified, although there were only one or two examples of many of them (Table 101 and Figure 506). Type A was used at this site for all definitely round-headed nails, plus those that were probably round-headed. Very few of the nails were cleaned, and it is not generally possible to tell from an X-ray whether a nail-head is round, oval or square. The other categories with flat heads of varying shapes (types B, J and R) are therefore under-represented. It is, however, possible to say that the bulk of the nails had flat heads of varying shapes - types A, B, J and R together represent 74% of the identifiable nails.

Two of the types of nail present are definitely post-medieval. Type O is a post-medieval horseshoe nail; both of the two examples must be intrusive. One is from a Period 2A pit (11316), the other from a Period 4-5 pit (14099). The single wire nail (Type X), which is a 19th-20th century form, was from a machining layer.

Hobnails are discussed with the leather footwear, in Function Category 1.

Table 101: Types of iron nail present at Elms Farm. The nail shafts are square and the heads are perpendicular to the shaft unless otherwise specified
Form Description Quantity % of identified nails
A Round, flat head 1979 56
B Square, flat head, often with rounded corners 166 5
C Round, mushroom-shaped head, hollow underneath. (The smaller forms of this are classified as hobnails, type G) 22 1
D Inverted triangular head, the same width as the shaft; in the same plane as the shaft 2 0
F T-shaped head, same thickness as shaft; square or rectangular shaft 5 0
G Hobnail; small mushroom head, short shaft 786 22
J Rectangular head, often with rounded corners 114 3
K Triangular head, same thickness as shaft; base of triangle to shaft, in same plane as shaft 22 1
M Sprig; headless nail 2 0
N Oval, domed head 2 0
O Horseshoe nail; cuboid head 2 0
P Clasp nail; oval or round head, with shaft distinctly offset from centre, usually located at one edge of the head 33 1
R Oval, flat head 341 10
S Rectangular head, slightly domed; rectangular shaft with wedge point 2 0
V Slightly expanded head, rectangular shaft 1 0
X Wire nail; round head, round shaft. 19th-20th cent. 1 0
BB Ball-headed nail or tack 3 0
FF Small L-shaped head, square-sectioned shaft, not cut 1 0
GG Hobnail with conical head 2 0
HH Slightly expanded head, square shaft 16 0
II Solid, truncated conical head with facetted sides 1 0
NN Inverted triangular head, same thickness as the shaft 1 0
PP Solid truncated conical head 1 0
SS Solid mushroom-shaped head, round or square shank 1 0
TT Truncated conical head with concave base 2 0

Structural and other nails

The remainder of the nails may be regarded as structural. Complete nails were relatively rare, forming only 19% of the assemblage (excluding hobnails). There were few large nails. Two samples of 100 type A nails from different areas of the site gave average lengths of 59mm and 57.6mm. In both cases, only one nail was greater than 100mm in length. Figure 507 shows the distribution of lengths from a sample of 100 from Areas K, H and J.

Figure 507: Plot of nail lengths for a sample of 100 Type A nails

For the site as a whole, only sixty-eight nails with a length greater than 100mm were recorded, some incomplete. The largest complete examples were a type B nail (square head) from a Period 3 layer in Area M, and a type TT (truncated conical head) from a Period 4-5 layer in Area K, both with a length of 170mm. There were no examples as large as, say, the largest from the Inchtuthil assemblage, which ranged from about 230-370mm in length, and which were clearly used for joining heavy timbers (Angus et al. 1962, 957). At Elms Farm, the lack of very large nails and relative scarcity of fairly large nails suggests that the use of nails with heavy structural timbers was rare on this site. Most of the nails could be from joined objects such as boxes and furniture.

There is, however, some evidence for structural use. The lining of boxed cremation 12203 may have been fixed to the sides of the cut using nails. Two sizes of nail appear to have been used in the construction; the larger ones have round heads with diameters of 20-23mm, with lengths of about 110mm, while the smaller nails have heads with diameters of about 15mm, and are about 90mm long.

There was also a group of nails from the temple foundation trench (5555, segment 5767, contexts 5512, 5562 and 5564) which appeared to be located at fairly regular intervals, and therefore might relate to the structure. Ten nails were collected from the segment, though not all were plotted, and only one was complete. The complete nail had a length of 79mm, not large enough to be a major structural nail, but potentially useful for nailing something to the wall. It was slightly bent at the point end. A bar fragment from the context (SF4726/4728) may be the shaft of a large nail at least 170mm long, which would make it the largest nail from the site. A second bar fragment (SF3233) is unlikely to be part of a nail. The plotted nails appear to be in the stake-holes within the trench, and, at the level at which they were found, would have been below ground level. It is possible that they are merely chance introductions into the fills of the stake-holes, either during construction of the building, or during demolition. None of the other excavated segments produced nails.

Iron bolts (not illustrated)

120. Bolt (holdfast), with circular head and rove (it is unclear which is which). The section is nearly circular. L. 52mm, diam of ends 36mm and 21mm, Diam. of shank 8mm. Fill 16328, Post-hole 16329, Group 4006, Area H, Period 4-5

121. Bolt or very large nail with a round, slightly domed head. The shank has no point, and may be complete. It has a right-angled bend, with the metal fractured most of the way through at the bend. This damage, which is ancient, may be the reason that it was discarded. L. 154mm, Head diam. 47mm. SF1468, 10000

Iron spikes

Figure 508: Fastenings and fittings, 123-232

122. Not illustrated. Spike, with a rounded head, point missing. Circular section, surface flaked. This is quite a stumpy object, and may be a tool rather than a 'structural' spike. L. 90mm, max. Diam. 23mm. Layer 6142, Group 543, Area H, Period 4.

123. Spike, pointed at both ends, square section. There are red corrosion products present on the surface. L. 219mm, max. W. 13mm. SF3985, Fill 10337, Ditch 10404, Group 838, Period 5-6

124. Not illustrated. Square section. The pointed end is complete, the other end possibly broken. L. 230mm, W. 11mm. SF762, Machining layer 4000

125. Not illustrated. Spike, with a regularly tapering rectangular section of constant thickness. The broad end was probably broken in antiquity. In good condition. L. 146mm, max. W. 16mm, Th. 9mm. SF4016, Layer 10372, Group 8014, Area F, not phased

126. Not illustrated. Head broken, square section. L. 124mm, max. W. 15mm. SF6635, Machining layer 15000, Area M

127. Not illustrated. Loop-headed bar, point probably missing. The taper suggests that this is a spike rather than the handle of e.g. a latch-lifter or key. L. 84mm, external diam. of loop 20mm, internal diam. 10mm. SF5854, Cleaning layer 5610, Area I, not phased

128. Loop-headed spike, curved. It is uncertain whether the curve is original. L. 81mm, external diam. of loop 14mm. SF5879, Cleaning layer 5598, not phased

129. Bar, ring-headed, complete as buried. The end of the rod is slightly curved, and appears rounded rather than broken. This is probably a loop-headed spike, but could be an incomplete handle from a lift key, as Manning O37. The rod appears to be tapering, which would be unusual for a key handle. The section is uncertain. L. 103mm, max. W. 6mm, external diam. of loop 18mm. SF6406, Fill 14381, Pit 14382, Group 260, Period 2B

Copper-alloy rivets

130. Not illustrated. Sheet fragment, rolled into a tight cone. This is probably the tip of a sheet rivet. L. 11mm. SF8368, Fill 14226, Pit 14225, Group 36, Area L, Period 2

131. Not illustrated. Part of a small domed rivet head, with no complete diameter. SF8382, Fill 20031, Pit 20030, Group 248, Area L, Period 2B

132. Not illustrated. Fragments, probably part of a rivet. In poor condition. c. 8x3mm. SF5259, 11269, Pit 11316, Group 227, Area N, Period 2A

133. Not illustrated. Thin sheet lozenge in fair condition, slightly holey. Probably an unused sheet rivet. L. 19mm, W. 9mm. SF579, Layer 5159, Group 457, Area J, Period 5-6

134. Rivet, with a solid, flattened mushroom head, cylindrical shank and damaged end disc. In poor condition, with little of the original surface surviving. Possibly military (see Crummy 1983, 118-19), although it is not certainly Roman. Diam. 11mm, L. 13mm. SF2487, Machining layer 4000 (Figure 508)

135. Not illustrated. Cast rivet. Slightly domed head, originally circular, edges damaged. Integral shank and circular rove. Ht 13mm, diam. 19mm, rove diam. 8mm. SF6505, Machining layer 11000


Iron (not illustrated)

136. Strip fragment, perforated. Probably a rectangular washer with one corner missing. 16x27mm, hole diam. 7mm. SF5119, Layer 13360, Group 600, Area I, Period 3B

137. Square washer with perforation. 40x40mm, hole diam. c. 9mm. SF1429, Machining layer 4000

138. Disc with a large central hole, broken on one side. Presumably a washer. 33x29mm, hole diam. 10mm. SF2302, Cleaning layer 5603, Area I, not phased

Lead roves or washers by Ros Tyrrell

These objects consist of a piece of lead sheet 2mm thick, through which a nail or rivet is passed to increase the size of the head. Six of these objects were found, all in machining layers. They may be post-medieval rather than Roman (not illustrated).

139. A roughly cut irregularly shaped piece of lead sheet with a hole 4mm in diameter. SF2725, Machining layer 4000.

140. Five more similar roves. SF6832, Machining layer 17000, Area Q.

Edge bindings

Copper alloy (not illustrated)

141. Edge binding with U-shaped section. In fair condition, surface flaked in places. Complete and almost straight, with one end flattened. L. 123mm, W. 7mm, Depth 7mm. SF698, Machining layer 4000

142. Straight edge binding strip with U-shaped profile. There is a possible rivet hole, in the centre of the U. One end is complete, the other incomplete and damaged. In good condition. Ht 4mm, internal W. 4mm, L. 84mm. SF1186, Fill 4380, Pit 4379, Group 741, Area K, Period 5

143. Edge binding strip, flattened, broken both ends. A narrow, straight strip, with a small broken iron object corroded on, probably a nail shaft. L. 51mm, W. 5mm. SF347, Machining layer 5000, Area J

144. Edge binding strip, U-shaped section, distorted. The length appears to be complete. There are no definite rivet holes. In poor condition, no surface surviving. W. 7mm, L. (straight) c. 110mm. SF5156, Cleaning layer 5603, Area I, not phased

145. Fragment of edge binding, with a U-section, both ends broken. In fair condition, some modern damage, bent. L. 60mm, Ht 8mm, W. across U 5mm. SF3395, Layer 6269, Group 581, Area H, Period 5-6

146. Sheet edge binding, with a U-shaped section. One end is complete as buried, and slightly distorted, the other damaged during excavation, with two small fragments broken off. The concave side is filled with white metal. XRF analysis showed that the binding was a quaternary bronze (cu/pb/sn/zn), while the fill was lead/tin (solder). L. 42mm, W. 6mm. SF8077, Fill 24398, Well 22210, Group 448, Area J, Period 6


147. Edge binding formed from two separate pieces of sheet. The first was a narrow strip formed into a U-shaped binding, with a slightly curved edge. It is unclear whether the ends are complete. A second, broader strip, which appears complete, was folded over this. The inner edge has two lugs with rivets through them. There is mineralised wood on both faces, at least some of which may be part of the original object. L. 63mm, W. of edge binding 8mm, W. of lugs 20mm. SF1988, Cleaning layer 4604, not phased (Figure 508)

148. Not illustrated. Edge binding, one end twisted and broken. It has a U-shaped section 5.5mm deep and 7mm wide. L. (straight) 98mm. Fill 15589, Pit 15594, Group 238, Area M, Period 2B

Double-spiked loops

Double-spiked loops are a very common find on Roman sites, and would have performed a variety of functions. The smaller ones were often used as staples for drop handles, particularly those made from copper alloy. Alternatively, two could be used together as a simple hinge. Those with everted points were probably used as suspension hooks. Four copper-alloy and twenty iron examples were found, in contexts from Period 2B onwards. Most of the iron examples had their points missing.

Copper-alloy (not illustrated)

149. The head has a sub-circular section, the spikes are square-sectioned and broken. W. across head 11mm, L. 30mm. SF2054, Cleaning layer 5307, Area J, Period 5-6

150. Distorted, made from a rectangular-sectioned strip of variable width. In poor condition, with little surface surviving. L. 39mm, W. across loop 13mm, max. section 6x2mm. SF5155, Cleaning layer 5603, Area I, not phased

151. Points missing, made from an ?oval-sectioned bar. L. 16mm, W. of head 9mm. Probably a casket fitting. SF4762, Cleaning layer 5609, Area I, not phased

152. Keyhole-shaped, made from a strip, max. W. 14mm. Tips missing. L. 27mm. SF3403, Cleaning layer 6552, Area H, not phased


153. Made from a strip 18mm wide. One point was broken off in antiquity, the other is bent outwards, and probably has its complete length, although the edge may be damaged. L. 102mm, W. across head 31mm. SF3499, Fill 10319, Unknown feature 10322, Period 3 (Figure 508)

Not illustrated

154. Points missing. L. 81mm, W. of head 18mm. Pit 23005, Group 236, Area N, Period 2

155. Small, complete, double-spiked loop. Probably made from a strip with a width of c. 7mm. L. 32mm, W. across head 15mm. Fill 14569, Pit 14574, Group 259, Area L, Period 2B

156. The ends are turned up, and the tip of one arm is missing. Possibly used as a hook. It was made from a strip which is thicker on the loop. L. 84mm, W. of loop 29mm, original L. (straight) c. 108mm. SF1980, Fill 4560, Post-hole 4559, Group 3085, Area K, Period 3

157. One point missing. Made from a strip, section 9x4mm, with the loop slightly thicker. L. 71mm, loop external diam. 27mm. SF4188, Prepared surface 4990, Area K, Period 3

158. Two joining fragments of a small double-spiked loop, complete but distorted. It was made from a strip, section c. 5x3mm. L. 33mm, W. across head c. 14mm. Prepared surface 5945, Area I, Period 3B

159. Points possibly broken. L. 104mm, external W. of loop 35mm, internal W. 15mm. SF1766, Layer 6053, Group 506, Area H, Period 3

160. Points bent and damaged. L 75mm, external diam. of loop 28mm, internal diam. 15mm. SF8052, Fill 23211, Pit 23125, Group 3047, Area N, Period 3

161. Points missing. Made from a rectangular-sectioned bar. L. 62mm, W. across loop 26mm, section 6x9mm. Fill 9244, Pit 9243, Group 806, Area D, Period 4

162. Points missing (old breaks). L. 49mm, section 7x4mm, W. of head 18mm. Layer 10104, Group 813, Area F, Period 4

163. Arms incomplete, made from a square-sectioned bar. L. 87mm, W. of loop 34mm. SF2194, Spread 10115, Group 813, Area F, Period 4

164. Object, very mineralised, and rather vague on the X-ray,. It is probably a double-spiked loop with the points missing. It certainly has a suitable shaped hole, but it is unclear whether it is a looped strip, or a bar with a perforated end. L. 34mm, W. 12mm. Fill 16073, Pit 16088, Group 559, Area H, Period 4

165. Made from a rectangular-sectioned strip. L. 67mm, W. of loop 27mm, section 12x6mm. SF6181, Fill 12237, Gully 12238, Group 971, Area R, Period 5

166. One arm was definitely broken in antiquity, and the other is probably incomplete. This is a heavy duty loop with a broad head. L. 74mm, W. across loop 30mm, section of loop 8x20mm. SF6439, Fill 15233, Pit 15232, Group 471, Area M, Period 6

167. Points missing. L. 63mm, external W. of head 32mm. SF765, Machining layer 4000

168. Both points missing, one arm bent. Made from a strip. L. 89mm, section 10x7mm, W. of head 32mm. SF1268, Machining layer 4000

169. The ends of the spikes are everted. Made from a strip, section 11x6mm. L. 86mm, head external W. 26mm. SF2119, Layer 5427, Group 8002, Area I, Not phased

170. Both spikes have recent breaks. Made from a square-sectioned bar. Loop external diam. 29mm, internal diam. 17mm. L. 37mm. , Machining layer 7000

171. Points damaged. Rectangular section c. 7x9mm, L. 75mm, W. across loop 30mm. SF1496, Cleaning layer 8000, Area E, not phased

172. Double-spiked loop with a broad rectangular-sectioned head, tapering either side to a thin, square-sectioned arm, one of which is broken. This appears to be an unusual form for a Roman loop, and it may be post-medieval. L. 118mm, W. across loop 32mm. Max. section of head 23x6mm. SF4297, Machining layer 11000

Iron staples

Staples and carpenter's dogs are very similar objects, both having a crossbar (usually with a rectangular section) with arms at each end. For the purposes of this report, flat-topped staples have the longer axis of the section of the crossbar at right-angles to the plane of the arms, and carpenter's dogs have the longer axis in the same plane as the arms. Flat-topped staples occur in contexts dating to Period 2 onwards.

Staples can also be U-shaped, but there are only two of this form in the assemblage. One is unstratified, and possibly modern, the other is from a Period 5 pit, but is not definitely a staple (Figure 508).

173. Flat-topped staple. It was made from a strip c. 9mm wide and 3mm thick, tapering abruptly at each end. L. 32mm, W. 23mm. SF6145, Pit 15818, Group 237, Area M, Period 2

174. Flat-topped staple, with a D-shaped crossbar, one arm surviving. The other end was broken off in antiquity. The shape is unusual. L. 49mm, max. W. 14mm, Th. 4mm, arm L. 19mm. Cleaning layer 5228, Area J, Period 5-6

Not illustrated

175. Staple or carpenter's dog, with squared top, one arm missing. The tip of the surviving arm is turned at right angles. The top is slightly curved perpendicular to the plane of the object. Top 74x7mm, arm L. 43mm. SF6992, Fill 11390, Ditch 11744, Group 2061, Area N, Period 2

176. Flat-topped staple. One arm is probably complete, and has been bent at an angle, with the tip everted. Arm L. 62mm, W.of top 76mm, section 8x4mm. SF7422, Fill 15816, Pit 15818, Group 237, Area M, Period 2

177. Flat-topped staple, with part of one arm missing. W. 47mm, arm L. 40mm, rectangular section c. 5x3mm. Fill 20196, Pit 20195, Area L, Period 2B

178. Flat-topped staple, made from a strip 5mm wide. W. 55mm, arm L. 37mm. SF7235, Fill 14706, Pit 20086, Group 36, Area L, Period 2A-B

179. Flat-topped staple, made from a strip, section 9x4mm. The arms are angled towards each other. L. of top 37mm, L. of arms 30mm. SF4386, Fill 9497, Ditch 9496, Group 150, Area D, Period 2B

180. Flat-topped staple, with one point missing. Square section. W. 37mm, arm L. 48mm. SF2817, Fill 8167, Well 8188, Group 788, Area E, Period 3

181. Object, probably a staple with a narrow, flat top, and long arms, one incomplete and the other almost entirely broken off in antiquity. Top 25x15mm, surviving arm length 64mm. SF5214, Fill 10393, Pit 10392, Group 361, Area F, Period 3

182. Probable flat-topped staple arm, complete, broken at the start of the crossbar. L. 63mm, max. section 4x4mm. Fill 5392, Pit 5394, Group 432, Area J, Period 4

183. Probable large staple fragment. An L-shaped bar, with one end pointed. The other arm is probably flat. There is a slight constriction at the bend. L. 71mm. Fill 10039, Pit 10038, Group 810, Area E, Period 4

184. Flat-topped staple, one arm complete. Section uncertain, probably square. W. 36mm, L. 32mm. SF5815, Fill 10922, Pit 10953, Group 673, Area N, Period 4

185. Two fragments of L-shaped bar, one lying nested inside the other. They are probably parts of the same flat-topped staple, with a rectangular-sectioned top and tapering points. Both pieces have fresh breaks to the crossbar. One arm is almost complete. Assuming that they are the same object, the crossbar would be >45mm long. Section 13x5mm, surviving arm L. 47mm. Fill 16073, Pit 16088, Group 559, Area H, Period 4

186. Flat-topped staple, with one complete and one half complete arm. Crossbar 30x8x5mm, arm L. 35mm. Fill 5229, Post-hole 5232, Group 427, Area J, Period 4-6

187. Flat plate, with a shallow curve. One end has a short arm, possibly complete, at right angles, set on one edge rather than in the middle of the end. The other end was probably similar. This may be a flat-topped staple or a large cleat, although it is rather thick. There is a small lump of ?lead in the corrosion, and multiple little flecks of the same material. L. 72mm, max. W. 18mm, Th. 9mm, arm L. 12mm. SF6429, Fill 15150, Pit 15231, Group 699, Area M, Period 5

188. Square-sectioned rod, probably a staple. It is bent into a loop with one arm longer than the other, but is possibly incomplete and somewhat distorted. L. 104mm, W. across loop 21mm. SF1379, Fill 8141, Pit 8142, Area E, Period 5

189. Strip, bent to form a spike at right-angles, other end broken. Probably a flat-topped staple. L. 72mm, max. W. 20mm, arm L. 37mm. SF5799, Fill 10891, Pit 10910, Group 676, Area N, Period 5

190. Cleat, or small flat-topped staple (X-ray is side view only). L. 27mm, arm L. 13mm. Fill 8737, Pit 8736, Group 902, Area P, Period 6

191. U-shaped staple, with a chisel point on the surviving point. Possibly modern. L. 45mm, W. 27mm. SF6589, Machining layer 4000

192. Probable flat-topped staple. L-shaped bar fragment, the end of the shorter arm pointed. The other end was twisted and broken in antiquity. Arm L. 75mm and 39mm, section 9x5mm. Layer 4540, Group 1163, Area K, not phased

193. Flat-topped staple, made from a strip. One arm is incomplete, but may have been shorter than the other one originally. Section 10x5mm, W. 32mm, arm L. 42mm and 27mm. SF4393, Fill 4842, Pit 4843, Group 1147, Area K, not phased

194. Large flat-topped staple, narrow in relation to its length. Rectangular sectioned throughout, and in fairly good condition. One arm has a fresh break, the other has flaked somewhat. L. 215mm, W. 85mm, section of top 25x13mm. SF7288, Machining layer 11000

195. Bar fragment, with one end turned at right angles. Both ends are broken. This is probably part of a flat-topped staple. The top has a constant section, probably rectangular, L. 94mm, W. 27mm. Arm L. 25mm. SF6404, Cleaning layer 15024, not phased

196. Probable flat-topped staple, broken at the top of the arms. L. 76mm, section 7x3mm. Cleaning layer 23002, Area N, not phased

Iron T-staples (not illustrated)

197. L- or T-clamp. It has only one surviving arm, and it is uncertain whether it ever had two. The part of the arm immediately above the shaft is broader than the arm itself, but not so thick. The end of the arm is missing. L. 66mm, arm L. 33mm, section 8x4mm. Fill 4842, Pit 4843, Group 1147, Area K, not phased

198. T-staple, with a lenticular head, slightly curved. The shaft is incomplete. L. 49mm, head 37x15mm. Fill 5162, Pit 5180, Group 443, Area J, Period 6

199. T-staple with a small lump of copper alloy in the corrosion. The shaft is incomplete, and has a right-angled bend. L. (straight) 75mm, W. across head 39mm. SF5856, Cleaning layer 5610, Area I, not phased

200. T-clamp, shaft broken, as Manning R68. Head W. 104mm. SF4377, Machining layer 7000, Area G, unstratified

201. T-clamp, with a straight cross bar with slightly tapered ends. Possibly complete. L. 80mm, head W. 81mm, section 7x5mm. SF4066, Fill 12026, Ditch 12027, Group 968, Area R, Period 4-5

Iron carpenter's dogs

202. Two carpenters' dogs, both in fairly good condition, and solidly made. They are very similar, with one end slightly rounded, although slightly different in size. The larger one is illustrated. a) L. 150mm, arm L. 62mm, section of top 21x14mm. b) L. 127mm, arm L. 47mm, section of top 21x14mm. SF702, Machining layer 4000 (Figure 508)

203. Complete bar the point of one arm. L. 105mm, section of top 17x c.10mm, L. of arm 50mm. SF2325, Prepared surface 6219, Group 516, Area H, Period 4

Not illustrated

204. Curved carpenter's dog, made from a bar with a pronounced taper to each end. The arms are of different lengths, and the shorter one has its point turned over. L. 72mm, L. of arms 23mm and 17mm. Max. section 9x7mm. SF406, Fill 4140, Pit 4139, Group 744, Area K, Period 6

205. Both arms are broken, though one is nearly complete. W. 52mm, section 6x4mm, max. arm L. 19mm. Fill 5375, Ditch 5359, Group 443, Period 6

206. Bar, with a rectangular section. It is curved, and was probably originally U-shaped. One end is broken, the other is broken at right-angles. This is probably an unusually long, curved carpenter's dog. A second, non-joining, fragment is probably the other end of the top, with part of its arm. L. 92mm, section 5x4mm. SF2087, Layer 5409, Group 457, Area J, Period 5-6

207. One arm is missing. Made from a square sectioned bar, W. 5mm. L. 57mm, arm L. 15mm. SF3198, Layer 5434, Group 1000, Area J, not phased

208. One arm missing. It is unusually long for the thickness of the crossbar, with very short arms. L. 90mm, section 7x5mm, arm L. 16mm. Fill 5472, Ditch 5473, Group 436, Area J, Period 5

209. One arm missing. L. 62mm, arm length 22mm, section 7x7mm. SF2747, Cleaning layer 5543, Period 5-6

210. Slightly curved lengthwise. Both arms are incomplete, although the longer one must be nearly complete, and is bent. L. 67mm, section 11x7mm, surviving arm L. 21mm. Cleaning layer 5543, Period 5-6

211. Carpenter's dog with a curved top. Both arms are broken. Cf. 5409 SF2087. L. 79mm, section c. 5x6mm. SF2969, Cleaning layer 5723, Area J, not phased

212. Arm from a carpenter's dog. Arm L. 32mm, section of top bar 10x4mm. SF2813, Layer 6008, Group 590, Area H, Period 5-6

213. Possible carpenter's dog fragment. A bar of constant width, broken, and tapering slightly to a 'tang' at right angles. L. 52mm, max. section 7x8mm. SF1454, Cleaning layer 8000, Area E, not phased

214. Broken across the top of the arms. L. 64mm, section 7x3mm. Fill 10039, Pit 10038, Group 810, Area E, Period 4

215. One complete, short, arm, other arm missing. L. 70mm, arm L. 22mm, section 11x8mm. SF2974, Cleaning layer 10236, Area F, not phased

216. Points missing. Top bar section 17x14mm, L. 126mm, arm L. 52mm and 48mm. SF5703, Machining layer 11000, Area A, unstratified

217. One point missing, the other incomplete. L. 158mm, section 19x10mm. SF5255, Fill 11247, Pit 11248, Group 207, Area N, Period 2

Iron collars

218. Collar, made from a strip of constant thickness, but variable width, which is an unusual feature. Diam. 52mm, W. 12-28mm, Th. 6mm. SF7654, Fill 16161, Post-hole 16162, Group 572, Period 5 (Figure 508)

219. Not illustrated. Collar, made from a strip of constant width. External diam. 55mm, W. 26mm. SF3230, Fill 5730, Pit 5736, Group 444, Area J, Period 6

220. Not illustrated. Collar, probably penannular, made from a strip of variable width. One end is complete, the other broken. External diam. 58mm, internal diam. 48mm, W. 26-37mm. SF908, Fill 6079, Post-hole 6080, Group 557, Area H, Period 4

221. Not illustrated. Curved strip in two joining pieces. Possibly part of an oval collar, c. 46x35mm. Section 9x6mm. Fill 10039, Pit 10038, Group 810, Area E, Period 4

222. Not illustrated. Collar, made from a strip, W. 19mm. Diam. 34mm. SF5859, Layer 13445, Group 600, Area I, Period 3B

Iron ferrules

223. Not illustrated. Conical ferrule, mouth broken. L. 75mm, max. diam. 30mm. SF1395, Post-pad 5166, Group 5006, Area J, Period 5-6

224. Not illustrated. Conical socketed ferrule. The section is uncertain, and is possibly square. Part of the socket was broken off in antiquity, but the length is complete. The socket is now almost triangular in section, but may be distorted. L. 114mm, max. Diam. 19mm, depth of socket 37mm. Fill 13499, Post-hole 13500, Group 3017, Area I, Period 3B

225. Ferrule? Conical, with a blunt point. The other end was broken in antiquity. It has a very wide gap, more of a channel than a socket, and it is difficult to see it functioning as a ferrule. However, it could be an unused example. It is similar in form to Manning S63, which is described as a ferrule, but also has a very open socket. L. 104mm. SF5928, Fill 15150, Pit 15231, Group 699, Period 5 (Figure 508)

226. Socketed ferrule, with a circular-sectioned socket, almost closed, and square-sectioned point, slightly damaged. L. 112mm. SF8419, Fill 4009, Pit 4008, Group 756, Area K, Period 3

227. Not illustrated. Cylindrical ferrule, made from a strip with the ends butted together. 24mm long, external diameter 16-19mm, Th. c. 1mm. Layer 15355, Group 701, Area I, Period 3B

228. Not illustrated. Ferrule. Circular collar, probably with two small tacks, through opposite sides. External diam. 17mm, Th. 4mm, Ht 16mm. Fill 8277, Pit 8280, Group 811, Area E, Period 4

229. Not illustrated. Spiral ferrule with three coils, formed from a strip 7mm wide. L. 35mm, external diam. 21mm. SF5370, Cleaning layer 5601, Area I, not phased

Hinge fittings

Hinge components appear to be poorly represented at Elms Farm. However, there may have been considerable use of hinges made from two double-spiked loops, as the illustrated example (no. 232), or simply interlocked. The group includes one of the largest iron objects from the site, a substantial strap hinge (no. 233), which must have been part of a large door.

Copper alloy

230. Pinned hinge, surface obscured by hard corrosion. There is no obvious attachment hole, although the object was not X-rayed. This is a robust little hinge, and may be post-Roman. L. 51mm, W. 14mm, Th. 5mm. SF922, Machining layer 4000, Area A, unstratified


231. Not illustrated. Loop-hinge fragment, slightly curved, probably from a box hinge. It is very similar to a hinge found in situ on a box from Colchester, which has a similar degree of curve at the loop end (Crummy 1983, 85, no. 2199). L. 50mm, W. 16mm, loop external diam. 16mm. SF5263, Cleaning layer 5543, Area J, Period 5-6

232. Hinge, formed from two double-spiked loops pivoting on a short rod with slightly flattened ends. The loops are now lying parallel. Part of one side is missing due to recent damage. L. 62mm, W. 27mm. SF2805, Cleaning layer 6000, Area H, not phased (Figure 508)

233. Not illustrated. Large drop-hinge. The hinge is a simple U-shape, 275mm long, and made from a strip with a rectangular section, 38x7mm, and slightly rounded ends. Unlike many drop-hinges, the arms are of equal length. The internal width between the two arms is c. 62mm, indicating that this hinge was from a substantial door. By comparison, the two large strap hinges illustrated by Manning (1985a, R8 and R9) both have gaps between the arms of about 35mm, even though one is considerably longer than the Elms Farm hinge. The X-ray was not located, so it is uncertain how many nail-holes there were. There appear to have been three, possibly four, nail-holes on each arm. SF1642, Fill 6126, Pit 6127, Group 577, Area H, Period 5

234. Not illustrated. Strip fragment, probably from a drop hinge. It tapers, with the narrower end complete and the broader end broken across a curve. There are two rectangular perforations. L. 82mm, W. 15-23mm. SF60, Fill 2117, Slot 2124, Group 3082, Area W, Period 3

235. Not illustrated. L-shaped hinge pivot, incomplete. Sections unclear. Spike, L. 70mm; pivot, L. 21mm. SF3226, Post-hole 5347, Group 173, Area J, Period 2B-3

236. Not illustrated. L-shaped hinge pivot, with a square-sectioned spike and a round-sectioned pivot, probably broken. Spike L. 85mm, pivot L. 42mm. SF761, Machining layer 4000

237. Not illustrated. L-shaped bar, possibly a hinge pivot. The sections are uncertain, although one arm appears to have a constant section, either square or circular, and the other tapers. Both arms are probably broken. Arm L. 47mm and 34mm, max. W. 15mm. SF5861, Layer 13348, Group 644, Area I, Period 5


Copper alloy

Figure 509: Fastenings and fittings, 238-252

238. Hook, made from strip of variable width, with a perforated terminal. Possibly a large steelyard hook. In two pieces, fresh break, poor condition, surface bubbled. L. 68mm, max. sect. 15x2mm, depth of hook 38mm. SF4666, Machining layer 11000

239. Hook, in two joining pieces. The stem is a circular-sectioned rod looped at one end, with the rod wound round itself. The other end broadens into a lenticular plate, with the tip coiled back on itself. This is possibly a steelyard hook, but if so, the hooked end must have been more bent originally. SF426, Fill 4187, Pit 4128, Group 744, Period 6


240. Wall hook; complete example with a U-shaped hook without a terminal knob. The spike is bent, which may be why it was discarded. L. 113mm, W. 45mm. SF7672, Cleaning layer 21615, Period 3B (Figure 509)

241. Not illustrated. Hook, possibly a Roman wall hook, although the straight element is rather short, and the curved element deeper than normal. L. 75mm, ht 68mm. SF769, Machining layer 4000, Area A, unstratified

242. Possible suspension hook. This object probably started life as a double-spiked loop, which was subsequently broken across the loop, and the remaining arm made into a crude loop. The section is variable. L. 55mm. Fill 4392, Pit 4429, Group 741, Area K, Period 5

243. Looped hook with everted points, made from a rectangular sectioned bar, section 13x9mm. This is a heavy duty hook which may have been part of a steelyard, or possibly part of a cauldron assembly. L. 69mm, L (straight) c. 107mm, W. across loop, 38mm. SF3006, Machining layer 11000, Area A, unstratified

244. Not illustrated. Large hook with a knobbed end, hook broken. Probably from a swivel fitting. L 155mm. In good condition, possibly Roman. SF784, Machining layer 4000, Area A, unstratified

245. S-hook, with a closed loop at one end, and the other end almost closed. The terminal at the open end is scrolled, and the other end may also be scrolled. L. 48mm. SF8256, Layer 5775, Group 600, Area I, Period 3B

246. Not illustrated. S-hook, one end broken, made from a strip, section c. 8x5mm. This was possibly originally a W-shaped hook. 45x35mm. SF3321, Fill 5841, Pit 5805, Group 444, Period 6

247. Not illustrated. Two-pronged hook with a flat cross-bar. The prongs have a regular and identical curve, making it unlikely that this is a staple. It is very similar to a double-armed hook from Danebury (Sellwood 1984, 370, 2.192); the arms are the same length, with an identical curve, though the Elms Farm hook is slightly wider across the top. Cross-bar section 7x3mm, L. 42mm; prong L. (straight) 92mm. Layer 23057, Group 236, Area N, Period 2

Iron strap fittings

248. Not illustrated. Strip, with a D-shaped section and one rounded end, and one square perforation. Probably strapping, or part of a hinge. It is bent, possibly due to forcible removal from its setting. L. 99mm, W. 38mm, hole W. 8mm. SF4143, Layer 4689, Group 4019, Area K, Period 4-5

249. Not illustrated. Strip. One end is rounded and has a large perforation, 11mm in diameter which is unlikely to be simply a nail hole. The strip is now somewhat distorted, but was probably originally flat for most of its length. There is a kink at the broken end, possibly originally a right-angled bend. This may be a hasp plate, as has been suggested for a somewhat similar strip from Hod Hill (Manning 1985a, 126, R5). 167x23mm. SF7888, Machining layer 12263, Area B

250. U-shaped binding strip in two pieces, one end broken. The other end has a roughly lozenge-shaped terminal, probably damaged, with a circular perforation. This is unlikely to be a strap hinge, as there is only a single perforation, but may have been a reinforcement strip from the edge of a substantial piece of timber. L. 145mm, complete strip L. 273mm, W. 13mm, W. of terminal 25mm, perforation diam. 9mm. SF2288, Cleaning layer 5543, Area J, Period 5-6 (Figure 509)

251. Not illustrated. Strap terminal, fresh break, tapering slightly to a broader, slightly rounded terminal with a circular perforation. L. 88mm, strap W. 22-19mm, terminal W. 26mm, hole diam. 6mm. SF4946, Fill 15006, Pit 15005, Group 701, Area M, Period 6

Other iron fittings

252. Strip fitting, broken. It was probably originally a squared U-shape with a flange at each end with a nail through, only one of which survives. It was possibly a handle, although it seems a little small, or it could be a guide bracket for a draw bolt. L. 74mm, strip section 16x5mm. SF1502, Cleaning layer 8000, Area E, not phased (Figure 509)

253. Not illustrated. Disc, in good condition. Diam. 58mm, Th. c. 10mm. Probably a door pivot base, as Manning, R15. The pit from which it comes is adjacent to the entrance to the temple, and this may therefore have formed part of that structure prior to its deposition in the pit. Fill 5144, Pit 5145, Group 430, Area J, Period 4

254. Not illustrated. Bar. One end has a right-angled bend, with a fresh break just beyond. The other end is bent at an angle, and is complete, and slightly rounded. This is probably part of rectilinear binding, or a staple. L. 73mm, section 11x5mm. Layer 10492, Group 838, Area F, Period 5-6


Internet Archaeology is an open access journal based in the Department of Archaeology, University of York. Except where otherwise noted, content from this work may be used under the terms of the Creative Commons Attribution 3.0 (CC BY) Unported licence, which permits unrestricted use, distribution, and reproduction in any medium, provided that attribution to the author(s), the title of the work, the Internet Archaeology journal and the relevant URL/DOI are given.

Terms and Conditions | Legal Statements | Privacy Policy | Cookies Policy | Citing Internet Archaeology

Internet Archaeology content is preserved for the long term with the Archaeology Data Service. Help sustain and support open access publication by donating to our Open Access Archaeology Fund.