PREVIOUS   NEXT   LESS DETAIL   CONTENTS   HOME 

Building Materials: The Tile

Peter Warry

Insula IX tegula and imbrex assemblage | Dating evidence for the 'House 1' sequence

Table of diagnostic Tegulae with their probable dating : PDF (10 KB)

Insula IX tegula and imbrex assemblage

The analysis of the manufacture and dating of the Insula IX tegulae draws extensively on the work in Warry 2006 and individual references have only been given where it is necessary to highlight specific issues. Inter alia, Warry 2006 shows that tegulae may be divided into four Groups (A–D) based upon the forms that their lower cutaways take. Group A tegulae are the earliest and Group D the latest.

All the tegulae from Insula IX appear to be of a similar red fabric presumably derived from the Reading clay beds, although the colour varies depending on the degree of firing. A single fragment in white 'Eccles' fabric was observed but no calcareous fabrics were present. The normal full range of firing states was noted. One tegula appeared to have been dipped in a dark red slip and two imbrices appeared to have had their upper surface covered in a white slip.

All the tegulae appear to be made in open-bottomed four-sided moulds and no inverted box mould manufacture was noted. Almost all the Group C cutaways showed evidence of the use of a cuboid mould insert for part forming this cutaway. Many of the Group D cutaways also showed evidence for mould inserts. Upper cutaways were generally knife cut, often with the horizontal slice being somewhat above the surface of the tile, leaving a residual element of flange.

Most tegulae are marked with signatures normally drawn with the finger. Some later tegulae signatures are produced with a comb but none of these have been observed at Silchester. The distribution of signature types from Insula IX is as below.

Number of semi-circles % of assemblage
One 33
Two 30
Three 13
Four 21
Other forms 4

Table of Tile-Makers' Signatures

The virtual absence of non-semi-circular signatures is typical of civilian production. The proportion of three and four semi-circle signatures is unusually high and may in fact understate the true proportion because sometimes the outer semi-circles are very faint. The Group A cutaways only appear with three or four semi-circular signatures. A typically sized civilian tile works would probably only have employed a single tile-maker. Signatures have been deemed to represent the separate marks of individual tile-makers, in which case the Silchester supplier was considerably larger than average, employing four tile-makers throughout the production. The location of the tile works is not known although there is nothing in the tiles themselves to suggest that more than one tile works was involved.

Sixteen tegulae were recovered, with complete lengths ranging from 386mm to 492mm, comprising 2 Group A, 11 Group C and 3 Group D. Three of these also had measurable upper breadths giving overall dimensions of 351 x 408mm, 361 x 472mm and 364 x 492mm; the first of these is Group D and the other two are Group C. Tegula dimensions tend to reduce through the Roman period so, as expected, the three Group D tegulae also had the shortest lengths. However, some of the Group C tegulae were longer than the Group A tegulae.

The length of these Group C tegulae is surprising. However, the much larger assemblage of complete tegulae retained in the Reading Museum from the Victorian excavations provides some explanation. Graph A plots the dimensions of these tegulae plus the three from the current Insula IX excavation. It can be seen that the Group C tegulae split into two very distinct groups, the group of smaller tegulae fit well with the expected progressive reduction in size through time but the larger tegulae are anomalous.

Graph A
Graph A: Dimensions of Group C tegulae from the Museum of Reading Silchester COllection and from Insula IX.

Two of the (larger) complete Group C tegulae from Insula IX were not flat but longitudinally convex, such that they would have formed an arch shape if placed on a level surface. Many of the complete Group C tegulae from the Victorian excavation were also convex. However, they divided neatly by length such that in Graph A almost all of the Group C tegulae in the cluster of larger tiles are convex while almost all the Group C tegulae in the cluster of smaller tiles are flat. This suggests that the larger convex Group C tegulae were a deliberate special manufacturing batch. In fact some 20% of all complete tegulae found in Britain are convex and these were almost certainly used to cover the extrados of vaulted roofs.

Graph B plots the length of the lower cutaways of the two clusters of Group C tegulae above, and it can be seen that the tegulae with the larger overall dimensions also have longer lower cutaways. These longer cutaways would have increased the extent of the overlap when the tiles were placed on the roof and as a consequence would have used 15% more clay to cover the same area as the smaller Group C tegulae. This was a substantial disadvantage when multiplied up into extra transport costs, thicker roof trusses and walls etc., and requires an explanation. The most obvious answer is that these convex tegulae were deliberately provided with longer cutaways in order to improve the overlap between the tiles when placed on vaulted roofs as this would have improved the robustness of the installation (see Warry 2006, 115, for a sketch of how this would have worked).

Graph B
Graph B: Distribution of lower cutaway lenghts of Group C tegulae based upon overall tegulae size.

The curvature of the convex tegulae from the Museum and Insula IX was such that they would have fitted a semi-circular vault with a span of around 4m. This was substantially less than the span of the Silchester Forum-Basilica, which was probably not vaulted anyway, so the most likely use would have been on a bath-house. No bath-house has yet been identified in Insula IX, which suggests that these tegulae were imported from elsewhere in the town.

Only two imbrices with complete lengths of 403mm and 405mm were recovered. These fall in the middle of the 365–417mm range of the much larger number of complete imbrices recovered in the Victorian excavations of the town as a whole. It is notable that most of the Insula IX tile spreads contained very few imbrices, which would suggest that these were not demolition deposits but material introduced from elsewhere. Either the deposits were simply surplus tegulae or, more likely, they were second-hand roofing material where the imbrices had been removed for some other use.

Dating evidence for the 'House 1' sequence[top of page]

The lower cutaway form allows tegulae to be dated and the suggested dating is given below (utilising subsequent data, the overlap between Groups B and C has been extended beyond that given in Warry 2006)

Based upon work on the Caerwent forum basilica (Warry in Brewer and Guest forthcoming), there is a possibility that the dating of the Group C cutaways can be further refined by reference to the length of the lower cutaway: longer cutaways being earlier and shorter ones later.

A listing of all the diagnostic tegulae recovered from Insula IX together with their probable dating based upon cutaway forms and a refined possible dating based upon cutaway length is given in the table. The analysis of the individual buildings is given below.

Period 2 Timber Building 1, Room 2

Only one diagnostic tegula of Group A from a possible occupation layer (context 5340). Dating AD 40-120.

Period 2 Timber Building 2, Room 6

One context (4559) containing two Group C tegulae which were part of a possible flat surface. The cutaway lengths of both tegulae were medium. Dating AD 125-260.

Period 3 Timber Building 4

Nine contexts containing 5 Group A, 3 Group B and 20 Group C tegulae. The Group C tegulae have cutaways of mainly medium or intermediate length suggestive of manufacture in the period AD 125-200. Easily the most significant of these contexts is 4758, which contained over half the diagnostic tegulae and is interpreted as a demolition deposit lying either side of a wall. The context consisted of 4 Group A, 1 Group B and 10 Group C tegulae, the latter all having cutaways of medium or intermediate length. It is surprising that a roof should be formed of assorted cutaway groups that not only represent very different periods of manufacture but would also fail to mesh together satisfactorily on a roof. If indeed these were all on the same roof together, then the best interpretation would be that the roof was originally comprised of Group A tegulae and subsequently repaired or modified using later styles. If this is correct, then this would suggest that the original structure was built sometime before AD 120, modified in the mid-2nd century and demolished some time after, say, AD 175.

Period 3 Masonry Building 1

A single Group C tegula from the second course of the flint wall (context 4592). The short cutaway length would suggest a date AD 200-260.

Period 3 Masonry Building 2

One Group B tegula from a clay levelling deposit (context 3375) and one Group C tegula from a possible post-hole (context 4563). The Group C tegula had a long cutaway, which could indicate it was produced shortly after AD 125.

Period 4 Masonry Building 3

Six contexts from within this building yielded 12 diagnostic tegulae, 2 of which were Group B, 9 Group C and 1 Group D. The lengths of the Group C cutaways revealed no particular trend. The greatest number of tegulae, including the Group D cutaway, come from context 1141, which is the residue of the robbed-out wall 1797. This could imply that the wall, and hence the building, was constructed after c. AD 240. However, much of the tile was contained within a soil matrix, clearly suggesting that the robber trench was refilled with material that need not necessarily have originated from the original wall. The Group D tegula is therefore more likely to provide a terminus post quem for the robbing activity.

However, if it can be argued that the majority of the discarded tegulae found in the robber trenches represented redeposited material from the original wall construction then, as the majority of this is Group C, it would argue for an original construction date after c. AD 125, and probably at least twenty years later to allow the broken tile to become available for such a secondary use.


 PREVIOUS   NEXT   LESS DETAIL   CONTENTS   HOME 

© Internet Archaeology URL: http://intarch.ac.uk/journal/issue21/4/finds_tile.htm
Last updated: Wed Sept 12 2007