SECTION 5.3: SOMERFORD KEYNES NEIGH BRIDGE FINDS REPORTS5.3.1 LATE IRON AGE AND ROMAN POTTERY
(Kayt Brown) 5.3.3 METAL SMALL FINDS (Hilary Cool) 5.3.5 SCULPTURE (Martin Henig) 5.3.6 WORKED STONE (Fiona Roe) 5.3.7 BRICK AND TILE (Leigh Allen) 5.3.1: IRON AGE AND ROMAN POTTERY by Kayt BrownIntroductionThe assemblage comprises 10182 sherds (100187g) of predominately Roman pottery, with a small quantity of prehistoric material, largely residual in Roman features. The main assemblage can be dated from the mid-late 1st century AD to the late 2nd century AD. A small number of late Roman shell tempered sherds suggest limited activity in the 4th century AD date, although no features were assigned to this date. ConditionAs with a number of the Cotswold Water Park sites, adverse soil conditions had a major impact on the condition of the assemblage; surface preservation was poor and many sherds displayed discolouring of surfaces hindering fabric identification. The average sherd size for the assemblage as whole was relatively low at 9.9g, although there was variation in sherd size between the phases. Evidence of use was represented by sooting on the exterior of vessels, post-firing holes in a number of vessels, sherds with rivet holes and a number of lead rivets (see Cool small finds, section 5.3.3). A total of 236 contexts produced pottery, although a number of large cleaning layers and unstratified material accounts for 37% of the assemblage by sherd count (36% by sherd weight). Ditches, layers and pits were the principal feature types to produce pottery although material was also recovered from gullies, postholes a well and a corndrier. MethodologyThe assemblage was recorded following the standard Oxford Archaeology recording system (Booth, OA unpublished). Sherds were examined by context and assigned to ware (or generalised ware group) and vessel type, where appropriate. Unstratified material and material from cleaning layers was recorded to generalised ware group as this material still provides an insight into the character of the assemblage. Where possible reference is made to existing typologies, for example Oxfordshire products (Young 1977), and fabrics are cross-references to the National Roman Fabric Collection (Tomber & Dore 1998). Sherds from each context were then quantified by sherd count, weight, rim count and rim equivalents. Also recorded were decoration, condition, evidence of use/reuse and repair. FabricsA quantification of fabrics by sherd count, weight and estimated vessel equivalents (EVEs) is presented in Table 1. Fine and specialist waresImported fine and specialist wares were represented by amphorae, samian ware and a single sherd of Lyon ware. British material comprised fine wares, mortaria and white wares. Amphorae were predominately body sherds of South Spanish Dr 20 olive oil containers, with a single handle sherd and a complete, if fragmentary rim. In Britain such material dates from the late Iron Age - 3rd century AD. Other amphorae were represented by a single body sherd of another south Spanish type, Cam 186, used for fish-based products and dated from the Flavian period to the early 2nd century AD. Samian ware products of the south Gaulish and central Gaulish industries (Les Martres-de-Veyre and Lezoux) were represented, but there was an absence of later east Gaulish material (see samian report below). Examples of British fine wares were restricted to two sherds of colour-coat wares, possibly North Wiltshire products but in poor condition, and a single north Wiltshire glazed ware sherd. North Wiltshire colour-coat material has been dated to AD 125-140/50 (Anderson 1978, 380-3), and the distribution of this ware is largely based on Wanborough and Cirencester, with isolated finds on sites across Wiltshire (Seager-Smith 2001, 240). The distribution of known sherds of North Wiltshire glazed wares is again centred on Wanborough and dated to the first-second century AD (Seager Smith, 2001, 253). Regional mortaria can be sourced to Cirencester and Oxfordshire (both white ware and white-slipped), with a small number of un-sourced oxidised sherds. Samian WareThe samian assemblage comprised 152 sherds (1299 g), representing a minimum of 45 vessels. Over half of this material was unphased (53% by weight, 50% by sherd count). Of the phased material, the context groups were very small, often only 1 or 2 sherds. The dating of contexts, and even features by the samian present, is therefore, in many cases likely to be unreliable given the likelihood of re-deposition and intrusion of sherds. South Gaulish material (La Graufesenque) accounts for just under a quarter (24% by sherd count) of all the samian recovered. Within this the earliest forms are rims from the bowl types Ritterling 12 and Curle 11. Production of Rit. 12 had ceased by AD 80, while Curle 11 developed through the Flavian period, and given the absence of earlier forms, it seems likely that samian started arriving at the site during the early Flavian period. This is similar to the situation at Claydon Pike (Webster FCP samian, section 3.2). Central Gaulish products form the bulk of the samian assemblage, dominated by Lezoux (post AD120), with only 25 sherds of Les Matres-de-Veyre (AD 100-125). The small quantity of Les Martres is interesting in that it shows that the acquisition of samian to the site did not decline at the end of the 1st century, alongside the decline of the south Gaulish industries. The proportion of forms Dr 18/31 to Dr31 within the assemblage is roughly equal, with marginally more Dr 18/31 present. This would indicate that the 2nd century assemblage is more likely to date to the first half of that century. Although form Dr 33 is the dominant cup form, there are a small number of Dr27, and the absence of any late 2nd century forms such as the samian mortaria form Dr 45, or forms Walters 72 and 79/80 would support this theory. Decoration occurs predominately on Dr 37 bowls (Figure 5.3.1: Decorated samian), and there are both south and central Gaulish examples. Decorated sherds comprise 15% of the assemblage, giving a ratio of decorated to plain ware of 1:10. This ratio is similar to that at Area A Asthall (Booth 1997, 110), and significantly higher than at the nearby site of Claydon Pike where decorated material accounts for only 4% of the samian assemblage (Webster FCP samian, section 3.2). There were three stamps, all on base sherds. Evidence of curation of samian vessels is visible on a number of sherds, in the form of rivet holes with in one instance a rivet still intact. Catalogue of illustrated Samian vessels (Figure 5.3.1: Decorated samian)
Coarse waresPrehistoric A small quantity of Iron Age sherds was present within the assemblage, all residual in later contexts. Fabrics were in the main calcareous, being either coarse shell or limestone tempered. These fabrics are comparable to those at Claydon Pike (Booth FCP pot, section 3.2) and Thornhill farm (Timby 2001, 24) and may indicate some degree of middle - late Iron Age activity either at, or near to the excavated area. A further four sand tempered, possibly middle Iron Age, sherds were also recovered. A small number of coarse flint-tempered sherds, and two quartzite-tempered sherds are likely to date to the Bronze Age or earlier, although these again occur alongside later material. Late Iron Age/early Roman The ‘belgic’ type wares, characteristic of the 1st century AD in this region, accounted for 17% of the assemblage (by sherd count, 19% by weight). The grog-tempered wares (E80) were the dominant fabric types, occasionally with varying amounts burnt/leached out organic material and shell, which may have resulted in some overlap between the E80, E10 and E40 ware groups. A small number of sherds in a flint-gritted fabric and a range of sandy wares were also recorded. A similar range of ‘belgic’ type fabrics were identified at Ashton Keynes (Wessex Archaeology 1989), Thornhill Farm (Timby 2001, 28) and Claydon Pike (Booth FCP pot, section 3.2). Roman The coarse wares were dominated by the reduced wares (43.3% of the total assemblage by sherd count). The condition of the assemblage precluded the identification of many sherds to a specific source or even industry, hence the majority of reduced wares were recorded in the general R30 category (62% by sherd count). Sourced material included products of the North Wiltshire industry (Anderson 1979, 9) which were being produced from the early 2nd century at known sites such as Purton Toothill Farm and Whitehill Farm. This material accounted for almost 4% of reduced coarsewares (1.7% of the assemblage by sherd count) although it is likely that many more sherds were recorded within the R30 category. Savernake and a possible variant fabric R38 (sand and grog-tempered) were marginally better represented. A similar pattern was observed with the oxidised coarsewares; North Wiltshire products accounted for 21% (by sherd count) of this group and less well represented but still significant were Severn Valley products and the coarse oxidised ware group O80. The unsourced categories again accounted for the majority of oxidised coarsewares, but included in this material are likely to be further products from the North Wiltshire and possibly Oxfordshire industries. As at Whelford Bowmoor, the reduced fabric R37 is again notable by its absence. A major fabric at Asthall (Booth 1997, 114), it does not appear to figure significantly at Whelford Bowmoor or Somerford Keynes, is poorly represented at Kempsford and no mention is made of similar fabric at Ashton Keynes (although to date no quantified data is available for this site). Again, this apparent paucity of R37 could be a result of the poor condition of much of this material, or be a genuine reflection of the distribution of this fabric. However, sherd condition is the more likely explanation, given that R38 is thought to be a variant of R37 from the same source. Black-burnished ware and imitation, possibly local, fabrics were, in common with a number of sites in the area, well represented comprising 14% ( by sherd count). The poor condition, however, of much of this material is reflected in the proportion by weight (g) of only 11%. Vessel form classes comprised Amphora (A), flagons and jugs (B), jars (C), jars/bowls (D), beakers (E), cups (F), bowls (H), bowls/dishes (I), dishes (J), mortaria (K), lids (L), and miscellaneous (M), with each of these classes sub-divided into specific form types. Table 2 shows the correlation between these vessel classes and the ware groups. Jar forms dominated the assemblage, accounting for 61% by eves. Within this class early forms include bead-rim jars (CH) (Fig. 5.3.3: Pottery from Phase 2, no.6), dating to the 1st century AD, occurring in the Belgic type fabrics and reduced coarsewares in roughly equal proportions. High-shouldered or ‘necked’jars (CE) (Fig. 5.3.2: Pottery from Phase 1, no.4) are similar in date, and again occur principally in the E ware fabrics, but are relatively poorly represented. The everted-rim cooking jar form (CK) (Fig. 5.3.4: Pottery from Phase 3, no. 18) occurs almost exclusively in Black-burnished ware and imitation fabrics and is the most prolific jar form within the assemblage. The only other jar form to be significantly represented is the medium mouthed jar (CD). A number of vessels were assigned to the intermediate jar/bowl category, when insufficient profile survived to enable a more precise definition. These comprised just over 11% by eves, as did bowl forms. Again black-burnished ware types would appear to be the main supplier of this form, mainly straight-sided bowls (HB) with either plain, slightly beaded or triangular beaded rims. There were a small number of plain top rims, dating from the early 2nd century AD and early flanged bowls, late 2nd to mid 3rd century in date. Fineware bowls occur as a single colour-coat sherd, a sherd of a Lyon hemispherical bowl (Greene 1979) (Fig. 5.3.2, no.5), and the samian forms Dr30 (HA), the curving sided (HC) bowl forms Dr31, 37, 38 and curle 11. Cup forms occur exclusively in samian forms Dr 27 (FB), 33 (FC), and 35 (FA). Alongside unsourced oxidised and reduced beakers are single examples of a poppyhead beaker and a decorated sherd of a north Wiltshire glazed ware (Fig. 5.3.3, no. 15), comparable to a conical beaker, part of the Wanborough group (Arthur 1978, 323, fig 8.8, no 5.5; Seager Smith 2001, 295, fig 102). Only two tankards were present, both Severn valley products (Fig. 5.3.3, no.10). Flagons are poorly represented, as are lids, mortaria and amphorae, with one complete, if fragmentary, rim of the latter. There is also a single example of an oxidised triple vase base (Fig. 5.3.4, no. 25) and a small fragment of a tazza in a reduced fabric. Catalogue of illustrated sherds (except samian)Figure 5.3.2: Phase 1 pottery 1. FS 3083. Jar. Handmade. Out-sloping jar rim, fabric L2, context 315/A/1 2. FS 3079. Jar. Handmade. Cordon at base of neck and groove on shoulder, fabric E40, context 314/A 3. FS 3157. Jar. Fabric R90, context 324/B/3 4. FS 3159. High shouldered jar, fabric E10, context 324/B/3 5. FS 486. Lyon ware, rim of hemispherical bowl, fabric F41, context 130/C/1
Figure 5.3.3: Phase 2 pottery 6. FS 3450. Bead rim jar. Fabric R90, context 400/A/5 7. FS 3333. Jar rim. Fabric R94, context 400/A/3 8. FS 3392. Jar rim. Fabric O30, context 400/A/3 9. FS 3401. Bowl. Fabric O40, context 400/A/3 10. FS 3318. Tankard with single handle. Fabric O40, context 400/A/2 11. FS 3395. Beaker rim. Fabric O30, context 400/A/3 12. FS 3402. Reed-rim bowl, Fabric O10, context 400/A/3 13. FS 3404. Bowl with spout. Fabric O10, context 400/A/3 14. FS 3461. Grooved flange bowl, fabric R30, context 400/A/5 15. FS 3530. Decorated sherd of North Wiltshire glazed ware. Fabric F22, context 407/C 16. FS 2251. Everted rim jar with faint burnished decoration on shoulder. Fabric E40, context 172 17. FS 2252. Straight sided dish. Fabric O80, context 172
Figure 5.3.4: Phase 3 pottery 18. FS 1908. Rim of angled everted rim jar. Fabric B11, context 164 19. FS 2527. Rim of curving sided bowl. Fabric B11, context 164 20. FS 2808. Dressel 20 amphora, context 252 21. FS 1924. Flagon/Jug. Fabric O30, context 164 22. FS 1973. Straight sided bowls. Fabric B11, context 164 23. FS 1969. Rim of angled everted rim jar. Fabric B11, context 164 24. FS 1968. Rim of angled everted rim jar. Fabric B11, context 164 25. FS 2648. Oxidised triple vase base. Fabric O20, context 227 The main forms of decoration comprised grooves and cordons on examples of most fabric types. Burnisheddecoration was present as either zones of burnish, lines or lattice. Lattice decoration was restricted to black-burnished ware types with only a few examples on reduced coarsewares. In all but one instance lattice decoration was narrow (acute), in keeping with the early date of the assemblage, and occurred on both jar and bowl/dish forms. Two vessels had burnished decoration on the underside of the base. Barbotine decoration and rouletting were the only other forms of decoration to occur on a number of reduced body sherds. Possibly the latest sherd in the assemblage was a rilled late Roman shell-tempered sherd, 4th century in date (unstratified). A large proportion of the assemblage was unphased (37% by sherd count). The bulk of the assemblage was recovered from trench 5, which is also the only area to produce any reliable phasing information. A brief discussion of the material from each trench is presented followed by comments on the overall assemblage. Although 3 broad phases were identified through the stratigraphy, in ceramic terms the distinction is not always clear. There is significant overlap in the wares represented in all phases, due partly to the narrow time span of activity at the site and longevity of some fabrics during this period, but re-deposition of sherds, and in some cases curation of vessels are also likely factors. The inter-cutting nature of many of the features to produce pottery, particularly in trench 5 has resulted in many features from different phases producing a quite homogenous range of wares, with dating, particularly between phases 2 and 3 based largely on a small number of diagnostic forms (Table 3). Most features within phase 1, including the enclosure ditches can be dated to the late 1st - early 2nd century. The ‘belgic’ type wares and early reduced coarsewares (such as Savernake) form the bulk of the material recovered in this phase. There is very little mortaria or samian and no British fine wares. Residual Iron age material amounts to 60 sherds. There are a few features which may indicate earlier activity at the site, although the individual assemblages recovered from these features are small. Posthole 310 contained grog-tempered sherds and limestone sherds, a combination that is indicative of the early to mid 1st century AD at the nearby site of Thornhill Farm (Timby 2001, 24). Ditches 117, 314, and gullies 315 and 316 also contained mid-late 1st century AD pottery. The ceramics from this phase 1 are comparable in both range of fabrics and forms, to Thornhill Farm periods E-F (c AD 75 - 120), which also appears to be a phase of intensive occupation. At Thornhill Farm, however, the quantity of ceramics diminishes during the 2nd century (phase 2, Thornhill Farm period G). At Somerford Keynes there is an increase in the amount of samian and black-burnished wares, including in the latter instance straight-sided bowls/dishes with flat top rims, dated from the early- mid 2nd century (phase 2). In trench 5 it was possible to further sub-divide this phase into phase 2a and 2b, on stratigraphic grounds, although again this is not reflected in the ceramics from these features. Included within phase 2 is the pottery recovered from the postholes of the aisled building (B1), which is consistently 2nd century in date (see below), with a small quantity of grog-tempered wares. Grog-tempered wares continue to appear alongside later fabrics into phase 3 and although some may be due to re-deposition, the average sherd sizes of this material remain high. Table 4 shows the pottery totals from the different trenches. Trench 1Single sherd of O20 (4g) and post medieval sherd (16g) from Ditch LB3. Phase 2/3. Trench 2A single black-burnished type sherd (7g) from gully 8, unphased. Trench 4Five sherds, comprising two R30 sherds (23g), two central Gaulish sherds (9g) dated to the early-mid 2nd century, and a single sherd of R95 (10g). Trench 5Trench 5 produced 7027 sherds (70070g), the largest assemblage of pottery recovered from a sequence of inter-cutting ditches, gullies, pits and post-holes. Good stratigraphic groups were, however, rare and many features produced small quantities of material. This, combined with the poor condition of much of the material, enabled only a broad phasing sequence to be applied within the trench (see above). Phase 1 Pottery from the enclosure ditches was predominantly late 1st to early 2nd century in date and all enclosures displayed a similar range of wares. Groups of material from individual features were small ranging from 7 sherds in enclosure 3 to 304 sherds in enclosure 4. Enclosure 4 is stratigraphically the earliest of the enclosures and was the only enclosure to produce a quantity of middle Iron Age sherds, re-deposited alongside later wares. Central gully 147, located within enclosure 2, also produced late 1st - early 2nd century pottery and could well be contemporary with this enclosure. A number of linear ditches were the other principal feature type to produce pottery (see Table 5), although again groups tended to be small and broadly dated to the late 1st - 2nd century. The only possible exception to this is posthole 311 (310), which contained grog-tempered and limestone sherds, which may indicate an earlier, mid 1st century AD date. The calcareous and e-ware fabrics are the two dominant groups within the phase 1 material from ditches, with e-wares remaining a consistent presence in all phases. Black-burnished ware and imitation types also feature significantly (see Table 5). Phase 2 The ditches assigned on stratigraphic grounds to phase 2 show an increase in the proportion of oxidised and reduced coarsewares and black-burnished ware compared to earlier fabrics (Table 6). Fine and specialist wares in the form of samian and mortaria also increase. However, material from ditch fills, given the nature of the feature type, should always be approached with caution, especially when dating is based on such relatively small samples. Also within phase 2 are a series of postholes from a substantial aisled building (B1), which produced 171 sherds of pottery within the limestone packing of the postholes. This material included 3 middle Iron Age sherds, a single calcareous sherd and a small quantity of grog-tempered wares, with the bulk of the material comprising roman reduced and oxidised coarsewares. Two sherds of Les Martres-de-Veyre samian, ( a decorated Dr 37 bowl and part of a DR 35 cup) would indicate a terminus post quem of 100 - 125AD for the construction of this building. Phase 3 There is very little in the assemblage which can be dated later than the late 2nd - early 3rd century AD (Table 7). With the exception of some later forms (for example groove and bead rim black-burnished ware bowls) and some mortarium forms, the main wares represented do not significantly alter during the 2nd century. Consequently the material from the phase 3 features does not appear to differ greatly from the preceding phase. For example ditch 114 is stratigraphically the latest feature within this trench yet in terms of the pottery recovered can not be distinguished from phase 2. The corndrier (167) produced a range of re-deposited late 1st and 2nd century fabrics with no diagnostic forms. Ceramics usually indicative of the 3rd century in this region are absent, for example east Gaulish samian, (and no late samian forms), and no material from the large late Roman industries such as Oxfordshire colour-coat products are present within the assemblage. Trench 7Eleven sherds comprising 10 E80 sherds (54g) and a single limestone tempered sherd (1g) from gullies 313 - 317, phase 1. Trench 8Single sherds of Black-burnished ware sherd (1g), flint-tempered E60 (7g), E80 (27g), two sherds of R30 (36g) and 3 R35 sherds (113g). Ditches 65 and 64, phase 2/3. Trench 9Single sherd of R38 from gully 62, phase 2a. Trench 12Single sherd of E80 (32g) and R30 (8g) from ditch group 36, phase 2/3. Trench 13A total of 137 sherds ( 1085g) was recovered from trench 13 (Table 8). Over 30 sherds were retrieved from cleaning layers and topsoil. Ditches 314 and 317 produced 8 grog-tempered sherds (75g), including part of a high shouldered or ‘necked jar, typically 1st century AD in date. Possible phase 1 /2 sherds comprise 4 E80 sherds (72g) and 4 R30 sherds (11g) from Gully fill (313). The remaining 91 sherds were from features phased to the 2nd century and include a number of sherds from the ditch which also produced the sculptural fragment. Trench 17Trench 17 produced the second largest assemblage of 2478 sherds (23,076g; Table 9). Little reliable phasing information was available from this trench, although in ceramic terms it is similar to trench 5. Unphased material accounted for 261 sherds. Only 4 sherds (88g) were recovered from features dated to phase 1, with a further 84 (723g) from phase 1 /2. By for the bulk of the assemblage was recovered form later features, including pit 405 and gully CG419, which, although stratigraphically late, contained a small quantity of 2nd century pottery. Layer 34, phase 2/3, produced over 580 sherds mostly 2nd century in date, with an average sherd weight of 9g. Included in this material are some large sherds of grog-tempered and reduced bead rim jars (average sherd weight of 35g), which would imply a date early in the 2nd century date, rather than later, as sherds of such sizes are unlikely to have undergone much post-depositional disturbance. Pit 400 produced an exceptionally large number of sherds (1299 sherds, 10596g). Black-burnished wares were well represented within this at over 100 sherds, including everted-rim jars, bowls and dishes. Once again the grog-tempered and other ‘belgic’ type wares are suprisingly well represented along with a coarseware bead-rim jar with particularly large individual sherd size (3 sherds 325g). The good survival of these 1st century fabrics and forms could well be due to an element of curation as a number of lead rivets, some with coarseware fabrics adhering were also recovered from the site, although mainly from unstratified layers. Well (402) produced a small quantity of 2nd century AD coarsewares and a single sherd of Les Martres -de-Veyre samian. Trench 19Trench 19 produced a small assemblage of only 204 sherds (2662g) with a good average sherd weight of 13g (Table 10). Just under half of this assemblage (48% by sherd count) was unphased, resulting largely from topsoil and cleaning layers. Material which could be phased was from predominately late ditches (phase 2/3). Only 5 grog-tempered sherds were recovered, the remaining assemblage comprising 2nd century material including central Gaulish samian, amphora and at least 2 mortaria, both Oxfordshire products (1 white-ware, 1 white-slipped). The pottery from this trench may represent activity from phase 2 onwards. General discussionThe small number of possible Bronze age and Iron age sherds hint at limited early activity in the area, with stronger evidence for activity at the site possibly from the early- mid 1st century AD and certainly from the late 1st century AD. The assemblage from Somerford Keynes shows many similarities to a number of rural sites within the region. As at Thornhill Farm and to a lesser extent at Claydon Pike there is a late 1st century - early 2nd century component of the assemblage which still comprises a significant proportion of ‘local’ grog-tempered wares. At Thornhill Farm, grog-tempered material was still a dominant fabric, occurring alongside Severn Valley and Savernake wares in period E-F (AD75-120+) (Timby 2001, 31). Although no quantified data exists for the assemblage from Ashton Keynes, it would appear that there is a similar range of material present during the late Iron age/early Roman period. The occurrence of limestone tempered fabrics is also well recorded at these sites and at a number of other rural sites such as Watchfield, Oxfordshire (Laidlaw 2001, 255), Groundwell Farm, Wiltshire (Gingell 1982,61), Kempsford (Biddulph, forthcoming) and Faringdon, Gloucestershire (Bryan in prep.). However, in contrast to Claydon Pike, Ashton Keynes, and to some extent Kempsford, activity at Somerford Keynes appears to cease in the late 2nd - 3rd century. Locally produced wares are the principal sources for the assemblage and in keeping with rural sites in the upper Thames Valley, the proportion of fine and specialist wares is low, at only 2.5% (by sherd count) and 5.5% (by weight - a higher percentage reflecting the presence of amphorae and mortaria sherds). Sites at Old Shifford farm, Standlake (Timby 1995, 129) and Gravelly Guy, Oxfordshire (OA in prep) both produced less than 1% fine and specialist wares, compared to the urban assemblage at Asthall where the figure is almost 7% during the same period (Booth 1997, 134). There is little evidence within the ceramic assemblage to indicate that it represents anything other than a rural, domestic assemblage, which would appear to be in contrast with the small find evidence (Cool small finds, this vol). However there are hints that the occupants at Somerford Keynes may have had access to more luxury items, for example the presence of a Lyon glazed bowl, although the occurrence of all fine wares is severely limited. Combined with this, characteristically Roman forms such as mortaria, amphorae and flagons are all poorly represented within the assemblage, and suggests that Roman culinary practises may have had little impact on the inhabitants of the site. Like many rural sites of this period jars and bowls form the dominant vessel types. The presence of sherds from a triple vase and a tazza are the only elements of the assemblage that may indicate any form of ritual activity, but given the number of sherds involved this is a rather tenuous link. BibliographyAnderson, A. S. 1979 The Roman Pottery Industry In North Wiltshire Swindon Archaeological Society Report No.2 Arthur, P. 1978 The lead glazed wares of Roman Britain In: Arthur, P. & Marsh, G. 1978 Early Fine Wares in Roman Britain BAR Brit Ser 57 Biddulph, E, in preparation ‘Pottery’ in P Booth and D Stansbie, A Roman rural landscape at Kempsford Quarry, Gloucestershire Booth, P. M. 1997 Asthall, Oxfordshire: Excavations in a Roman ‘Samll Town’ Thames Valley Landscapes Monograph No.9, Oxford Archaeological Unit Booth, P Oxford Archaeology Roman Pottery Recording System: An Introduction Oxford Archaeology unpublished Bryan, E, Brown, K and Barclay, A, in preparation The Pottery, in J Cook, E B A Guttmann and A Mudd, Excavations of an Iron Age site at Coxswell Road, Faringdon, Oxon Gingell, C. 1982 Excavation of an Iron Age enclosure at Groundwell Farm, Blunsdon St Andrew, 1976-7, Wiltshire Archaeol Mag 76, 34-75 Greene, K. 1979 The Pre-Flavian Fine Wares at usk University of Wales, Cardiff Laidlaw, M, 2001 Pottery, in Excavations at Watchfield, Shrivenham, Oxfordshire, 1998 (V Birbeck), Oxoniensia 66, 2002, 250-265 Seager-Smith, R. The Pottery In: Ashton Keynes, Wessex Archaeology post-excavation assessment, unpublished Seager Smith, R 2001 The coarse pottery in: A S Anderson, J S Wacher and A P Fitzpatrick The Romano-British ‘Small Town’ at Wanborough, Wiltshire, Excavations 1966-1976 Britannia Monograph Series No. 19, pp 232-344 Timby, J, 1995 Pottery, in Iron Age and roman settlement at Old Shifford Farm, Standlake (G Hey), Oxoniensia 60, 1996, 124-136 Timby, J. 2001 The Pottery In: D. Jennings, J. Muir, S. Palmer and A. Smith Thornhill Farm, Fairford, Gloucestershire: an Iron Age and Roman Pastoral site in the Upper Thames Valley Oxford Archaeology Thames Valley Landscapes Monograph Tomber, R and Dore, J, 1998, The National Roman Fabric Reference Collection: a handbook, MoLAS monograph 2 Young, C J, 1977 Oxfordshire Roman pottery, BAR Brit Ser 43, Oxford 5.3.2 THE COINS By Cathy KingSomerford Keynes in archaeological terms is a rural site characterised by rectangular enclosures and a large aisled building which may have been an official tile depot, located approximately 5 km south of Cirencester. Two hundred and seventy-eight coins were recovered from the site ranging in date from the Iron Age and Roman Republic to the late fourth century AD implying a period of continuous occupation. (Table 1) Unfortunately most of the coins are unstratified and a substantial number are in such poor condition that they cannot be given a precise identification. Despite this limitation, many coins can be assigned within a date range narrow enough to determine an overall pattern of coin loss and set the SKNB material in a broader context of comparable finds defined chronologically, functionally, and geographically. One of the more interesting aspects of the group of coins from Somerford Keynes is the relatively high proportion of early coins as there are 95 coins (34%) that belong in the years from approximately 40 BC to AD 192. Of these, thirteen are British Iron Age pieces dated to between c. 40 BC and 30 AD and one is a Republican denarius of 32-31 BC. There are 31 coins from the first century minted between AD 36 and AD 96 including two genuine denarii together with the core of a plated one and 24 coins minted between AD 96 and AD 192 including three plated denarii and 21 bronze pieces. The 26 unidentifiable bronze coins minted during the first or second century AD cannot be assigned a closer date. This influx of early coin is relatively unusual on rural sites and may be related to the Somerford Keynes’ proximity to Cirencester, the significance of which will be discussed further below. The Iron Age coins are all Dobunnic or copies of Dobunnic silver units with one exception, a debased British LZ stater. There are nine uninscribed pieces, three of which are plated that belong in the late first century BC and three with the legend ANTED, two of which are plated, that are currently dated to the early first century AD. The predominance of Dobunnic Iron Age coins at this site is unsurprising since they occur frequently in Gloucestershire, Wiltshire, Oxfordshire, Hereford and Worcester, and Avon and more sporadically further afield including outliers in Essex and Kent. (Van Arsdell and de Jersey, 1994, 73-83) More problematic is the question whether these coins can be related directly to the Iron Age occupation of the site or whether they reached it in the early Roman period of occupation. Dobunnic silver while clustering in Gloucestershire, Oxfordshire, and Wiltshire had a wide distribution throughout Britain as noted above and virtually all finds turn up in post-conquest contexts. (L. Sellwood, 1984, 203) The presence of pre-conquest silver on Roman sites in Britain is comparatively rare apart from those which are both early and/or military in nature (eg Hod Hill, Alchester, Cirencester), temples (eg Hayling Island, Harlow), and civilian sites with a military supply component like Fishbourne. (Table 4) The single Republican denarius is the only silver coin recovered from Somerford Keynes until the Flavian period when two genuine denarii of Vespasian occur as well as a plated piece datable to AD 69-AD 96. It is unlikely that the Republican denarius reached the site before the conquest. Early bronze coins minted before AD 44 or copying coins minted before AD 44 again tend to be comparatively rare on most British sites which normally produce a higher proportion of bronzes from the Flavian period. (Table 5) Somerford Keynes has a single genuine as of Gaius, a group of eight post-conquest coins of Claudius I of which five (62%) are contemporary copies and three bronze coins of Nero. The high proportion of Claudian copies is typical of British sites yielding bronze coins minted before the Flavian period. The absolute numbers of the pre-conquest and immediately post-conquest SKNB coins is small but, by analogy with other unarguably early sites, their presence together with that of the Republican denarius seems to suggest the relatively early arrival of these coins at Somerford Keynes. At some sites the high incidence of bronze issues may be related to the invasion itself. Sauer has argued, for example, on the basis of the coin loss pattern of imperial bronzes minted before AD 41 found in Britain that a large number of coins of Caligula from military bases can be linked to their foundation date in the 40s and 50s (Sauer, 2000, 49). Despite the relatively large amount of early coinage, there is no direct evidence of military presence at Somerford Keynes in these years and the earlier of two groups of military small finds from the site has been dated to the later second and early third centuries AD (Cool, section 5.3.3). However, as noted above, Somerford Keynes is not far from Cirencester where a fort was established in the first century AD although it does not seem to have survived beyond the sixties. There are similarities in the coin loss pattern for bronze coins but less for the silver which may in part be related to the small sample size (34) of the Cirencester coins (Tables 4 and 5, no. 2). (Reece, in Wacher and McWhirr, 1982, 86) The coins from the town of Cirencester, a civitas capital, which were published in 1998 also have early pieces (Tables 4 and 5, no.1) including denarii of Augustus and Tiberius and a plated denarius of Claudius I. The bronzes include four coins of Agrippa minted under Gaius, 57 of Claudius I of which 13 are genuine and 44 (77%) are imitations, and fourteen genuine pieces of Nero. (Reece in Holbrook, 1998, 270) The general pattern of loss is the same as that for Somerford Keynes, but the percentages of the 1998 Cirencester coins are much smaller than those at SKNB since the sample size is much larger (3372) and the individual sites within the town have been amalgamated into a single group. The proportion of bronze coins recovered at Somerford Keynes (Table 3) in the Flavian period AD 69-AD 96 is higher than that of the preceding period and sixteen pieces, all genuine can be attributed to these years. That they all necessarily arrived at the site in these years is less likely since both first and second century bronze coins continued in circulation long after they were minted and it was only the cessation of their minting in the 260s AD combined with the development and subsequent reform of the silver coinage that led to their disappearance in the later third century. The percentage of coins of second century date (AD 96-192) also remains relatively high (8.5%) at Somerford Keynes (Tables 2 and 3). Four denarii, all plated, were recovered: one each of Trajan, Hadrian, Pius, and Commodus. Twenty bronze coins can also be securely dated to the second century spanning the period from Trajan through Commodus. The incidence of plated denarii while interesting is not unexpected since these are just the sort of coins that tend to be abandoned in ancient times once they have been rejected as false. The high number of illegible coins that cannot be attributed securely is frustrating but it is safe to argue that some at least belong in the second century A.D. Bronze coins of the third century A.D. are extremely rare on British sites and none datable to the period AD 192 to AD 260 have been recovered from Somerford Keynes. There are, however, two denarii from these years: a genuine piece of Julia Domna and a plated copy of Julia Mamaea. The number of coins recovered from Somerford Keynes that were minted between AD 260 and AD 402 is much higher than those datable to the years before AD 260 and in this respect the site conforms to the general pattern of loss on British sites in the later period. Within these chronological parameters, however, there are periods when coin loss peaks: AD 260-296, AD 330-348, AD 348-360, AD 364-378, and AD 388-402. While most British sites reflect a significant rise in the number of coins lost in the later third century, they do not display uniform patterns of loss in the fourth century. Reece, having noticed this phenomenon has attempted in recent years to examine whether this variation is linked to the nature of the site (eg villa, temple, small town, etc.) or its geographical location in the east or the west of the country. (Reece 1991, 1993, 1995, 1996) Lockyear, using the same data, has suggested some modifications to the methodology employed by Reece, and has been able to confirm that different types of sites seem to have different loss patterns although geographical distinctions do not emerge clearly. (Lockyear, 2000, 418-419) At Somerford Keynes the percentage of coins minted between AD 260 and AD 296 is higher at 19% than the years AD 330 to 348 where the percentage drops to 13% rising to 16% in the years AD 348 to 364 and dropping off sharply thereafter. The periods of peak coin loss between AD 260 and AD 364 can have high numbers of imitations as well as genuine pieces but the number of imitations of bronze coins is virtually non-existent after AD 364 on British sites when any official and deliberate addition of silver to the alloy has been abandoned. During certain periods of peak loss the numbers of imitative coins significantly elevate the level of coins recovered. At Somerford Keynes the percentage of official (27) and imitative (27) radiates is the same. However, there are twice as many genuine pieces (24) minted between AD 330 and AD 348 as there are copies (12). For the years AD 348 to AD 364, on the other hand, the copies (32) outnumber the genuine pieces by more than two to one. The proportion of imitations to the prototypes they copies is not uniform at all sites and thus far no clear pattern has emerged in terms of site type or geographical differences. In part this may be due to the fact that Reece has not chosen to focus on the proportion of genuine versus imitative copies from AD 260 onwards in his analysis of the pattern of coin loss in Britain, arguing that the definition of what constitutes a copy is subjective and that the criteria will vary from researcher to researcher. While he has a valid point, it should be clear from number of imitative coins dating to peak periods of coin loss in the later empire that they are a significant factor in influencing the pattern of loss itself within certain periods. Somerford Keynes is an interesting and somewhat unusual rural site in producing so much coinage from the years before AD 192 suggesting some sort of activity dating from the first century AD. In this aspect the coinage mirrors the picture provided by the small finds assemblages but contradicts that of the pottery. It should be noted that the high proportion of early coins will inevitably have the effect of depressing to some extent the percentages of coins from the periods of peak coin loss between AD 260 and AD 402 as does the number of coins from the third and fourth centuries that cannot be assigned to a specific period. Nonetheless the overall pattern at SKNB is similar to that of the whole of British sites. In this context if may also be worth noting that the latest building activity on the site is datable to the later second century. It remains to determine the relationship of Somerford Keynes to other British rural sites in general and to the other Cotswold Water Park sites more specifically, both of which will be examined in more detail below. BibliographyAllen, T.G. et al., 1993, Excavations at Roughground Farm, Lechlade, Gloucestershire: a Prehistoric and Roman Landscape, Oxford Booth, P.M. 1993, ‘Ducklington, Gill Mill’, Oxford Archaeological Newsletter, 18/3, 19-23 Booth, P.M., 1997. Asthall, Oxfordshire: Excavations in a Roman ‘Small Town,’ Thames Valley Landscapes Monograph 9, Oxford Booth, P.M. and Evans J., 2001. Roman Alcester, Northern Extramural Area 1969-1988 Excavations, Roman Alcester Series iii, CBA Research Report 127, York Brown, A.E. (ed.), 1995. Roman Small Towns in Eastern England and Beyond, Oxford Burnham, B.C. and Wacher, J.S., 1990. The ‘Small Towns’of Roman Britain, London Cook, P.P.M., 1955. ‘A Roman Site at Asthall, 0xfordshire’, Oxoniensia, xx, 29-39 Cracknell, S. and Mahany, C., 1994. Roman Alcester: Southern Extramural Area 1964-1966 Excavations. Part 2: Finds and Discussion. CBA Research Report 97, York Cracknell, S. (ed.), 1996. Roman Alcester: Defences and Defended Area. Roman Alcester Series ii, CBA Research Report, 106, York Cunliffe, B.W., 1971. Excavations at Fishbourne 1961-1969, ii, The Finds, Reports of the Research Committee, Society of Antiquaries, 27, London Drew, C.D., 1931. ‘The Excavations at Jordan’s Hill Weymouth 1931’, Proceedings of the Dorset Natural History and Archaeological Society, 53 Ellis, P., 1984. Catsgore 1979. Further Excavation of the Romano-British Village, Western Archaeological Trust, Monograph 7, Gloucester Faulkner, N., 1994. ‘Later Roman Colchester’, Oxford Journal of Archaeology, 13,1, 93-120. Hands, A.R., 1993. The Romano-British Roadside Settlement at Wilcote, Oxfordshire, i, Excavations 1990-1992, BAR British Series 232 Hands, A.R., 1998. The Romano-British Roadside Settlement at Wilcote, Oxfordshire, ii, Excavations 1993-1996, BAR British Series 265 Holbrook, N., (ed.), 1998. Cirencester, The Roman Town Defences, Public Buildings and Shops, Cirencester Excavations v, Cirencester Hurst, H.R., 1985. Kingsholm. Excavations at Kingsholm Close and other Sites with a Discussion of the Archaeology of the Area, Gloucester Archaeological Reports, i, Gloucester Leach, P., 1998. Great Witcombe Roman Villa Gloucestershire. A Report on Excavations by Ernest Greenfield 1960-1973, BAR British Series 266, Oxford Leech, R., 1982. Excavations at Catsgore 1970-1973, Western Archaeological Trust Monograph 2, Bristol Lockyear, K., 2000. ‘Site finds in Roman Britain: A Comparison of Techniques’, Oxford Journal of Archaeology, 19, 4, 397-423 Mahany, C.M. (ed.), 1994. Roman Alcester: Southern Extramural Area. Part I, Stratigraphy and Structure, Series i, CBA Research Report 96, York O’Neill, H., 1957. ‘Akeman Street, Quenington, Gloucestershire’, Transactions of the Bristol and Gloucestershire Archaeological Society, lxxvi, 33-43 Ralegh Radford, C.A., 1972. ‘Excavations at Cricklade’, Wiltshire Archaeological Magazine, 67, 61-111 Reece, R., 1980. ‘Religion, Coins and Temples’, in W. Rodwell (ed.), Temples, Church and Religion in Roman Britain, BAR British Series 77, i, 115-128, Oxford Reece, R., 1991. Roman Coins from 140 Sites in Britain, Cirencester Reece, R., 1993. ‘British Sites and their Roman coins’, Antiquity, 67, no. 257, 863-869 Reece, R., 1995. ‘Site-finds in Roman Britain’, Britannia, 26, 179-206 Reece, R., 1996. ‘The Interpretation of Site Finds – a Review’, in King, C.E. and Wigg, D.G., Coin Finds and Coin Use in the Roman World, The Thirteenth Oxford Symposium on Coinage and Monetary History, 25-27-3-1993, Berlin, 341-355 Richmond, I., 1968. Hod Hill 2, Excavations Carried out Between 1951 and 1958 for the Trustees of the British Museum, London Sauer, E., 2000. ‘Alchester, a Claudian ‘Vexillation Fortress’ near the Western Boundary of the Catuvellauni: New Light on the Roman Invasion of Britain’, The Archaeological Journal, 157, 1-78 Smith, R.F., 1987. Roadside Settlements in Lowland Roman Britain, BAR British Series, 157, Oxford Taylor, M.V. and Collingwood, R.G., 1924. ‘Roman Britain in 1923’, Journal of Roman Studies, 12, 1924, 240-287 Timby, J.R., 1998. Excavations at Kingscote and Wycomb, Gloucestershire: a Roman Estate Centre in the Cotswolds with a Review of other Nucleated Settlements in the Region, Cirencester Van Arsdell, R.D. and de Jersey, P., 1994. The Coinage of the Dobunni: Money Supply and Coin Circulationin Dobunnic Territory, Oxford Wacher, J.S., and McWhirr, A., 1982. Early Roman Occupation at Cirencester, Cirencester Excavations i, Cirencester. Wedlake, W.J., 1958. Excavations at Camerton, Somerset, Camerton Wedlake, W.J., 1995. The Excavation of the Shrine of Apollo at Nettleton, Wilts. 1956-1971, Report of the Research Committee, The Society of Antiquaries, 40, London Williams, R.J. and Zeepvat, R.J., 1994. Bancroft: The Late Bronze Age and Iron Age Settlements and Roman Temple-mausoleum and the Villa, vol i, Excavation and Building Materials, vol. ii, The Finds 5.3.3 THE SMALL FINDS by Hilary CoolIntroductionThis report deals with all the small finds from Somerford Keynes other than coins and those made of stone, and a few preliminary remarks about the nature of this material are appropriate before it is described in detail. The archive consists of just over 1000 relevant items all of which have been inspected. Only 13% of this total came from the excavation, the remainder was the result of surface collection and metal detecting. No detailed evidence of precisely where much of the surface collected material came from appears to have been recorded, and this naturally limits to some extent the questions that can be asked of it. The methodology adopted was to inspect and catalogue appropriately all the items. Where possible they were assigned a spot date. It became apparent that by far the majority of the identifiable items were of late Iron Age to early Roman date. There was a little mid 2nd and 3rd century and a slightly larger amount of late 4th to 5th century material. Medieval material was very rare. Post medieval and modern material consisted of the sort of items to be expected from sporadic agricultural use of the ground (horseshoes, horse fittings, buttons and other dress accessories). The following report discusses all of the material that can be typologically assigned to the Iron Age and Roman periods, together with a few items of intrinsic interest which may be Roman but for which no dated comparanda has been found. All of the material from the excavations is included but the less diagnostic material from the surface collection has been excluded. The collection is biased in several ways. The use of metal detectors has been shown to bias the metalwork assemblage, for example towards more ‘chunky’ items at the expense of long thin things, towards bow brooches at the expense of penannular brooches (Cool in Booth forthcoming). It is clear as well that on this site bone artefacts are seriously under-represented. It could be expected that they would be under-represented in the surface collection but even in the excavated material they are very rare with only two items recorded out of the total 130 stratified finds. Bone was the plastic of the Roman world, being a cheaper alternative to metal for many artefacts. On sites where bone is not well preserved, one can expect to be loosing many parts of the material culture of the inhabitants. It can also be expected that items that were made of iron will also be under-represented. The iron from the site came from both the excavations and survey, but has suffered considerable post-excavation deterioration in the intervening years. As it was not X-radiographed until 2003 as part of this programme of work, there was no record of what many of the items may originally have been, as they now exist only as formless fragments. With an assemblage subject to these biases, there are limits on the types of questions that can be asked of it. It should, and does, provide very useful dating information. The functional analysis may provide clues as to the status of the site, but it has to be appreciated that some functional categories will be under-represented. Some types of toilet implements and personal ornaments will be missing or rare because of the way the material was collected. Some categories such as writing equipment, tools and agricultural items can be expected to be reduced in numbers because they would normally consist of large numbers of iron items. The contribution of the small finds to a detailed understanding of the excavated trenches is also limited because of the small numbers found stratified; the fact that 17% of the excavated material consisted of relatively featureless fragments of wire, sheet; and the lack of knowledge of how the unstratified material relates to the excavated trenches. Despite all of the problems, however, the small finds do tell a most remarkable story especially when compared to the evidence of the pottery and glass vessels. Both of those categories suggest a modest rural assemblage. As may be seen by a glance at Table 1 the sheer quantity of small finds found suggests something very different. In what follows the material will be discussed first by functional category following Crummy (1983). The discussion will date the material and where possible set it in a wider context. This detailed consideration will be followed by a brief overview drawing out some important themes. A full consideration of what the small finds tell us about the site will be reserved for the overall consideration of all the small finds from all sites in this project, as it is only by considering the Somerford Keynes assemblage against a wider regional background that it becomes possible to understand it. It should be noted that the condition of the copper alloy was variable. Some was in very good condition, some in very poor condition and many items were so thickly coated by iron pan that only the use of a magnet demonstrates that they were made of copper alloy rather than iron. It has not always been possible, therefore, to describe all of the decoration accurately. Personal OrnamentsThe personal ornaments are summarised by phase in Table 2. The category is dominated by brooches to a quite an extraordinary extent. As these provide valuable dating evidence they will be discussed at length grouped according to a broadly chronological order. Brooches3rd to 1st century BC form
The earliest brooch found was the involute brooch 321. Brooches such as this are an Iron Age form in use during the 2nd and 1st century BC and were probably in existence by the end of the 3rd century BC (Harding 1974, 188-9). They had a variety of hinge mechanisms (see Hattatt 1985, 16-7) but the precise method used on this is unclear. An X ray shows only very slight hints that the corroded hinge plate may have been perforated. The expanded plate over the catchplate on these brooches was sometimes decorated, but there is no evidence that it was in this case though any light incisions could have been obscured by the corrosion. Early to mid 1st century forms
All the one piece brooches with an internal chord in the late La Tène tradition will be termed Nauheim Derivatives here (for general discussion see Olivier 1988, 36-8). In general a date prior to c AD 75 can be suggested for these. Simple wire examples known as Drahtfibel are represented here by 984, 5016, 5128 and probably by the fragments 149, 467 and 630. Most have plain bows but one (149) has transverse moulding which Olivier (1988, 37 no. 14) has noted as a rare but widely scattered variant. 984 also has traces of decoration. True Nauheim Derivatives have rectangular bows, and are more common at Somerford Keynes than Drahtfibel. Seven examples have decorated bows (148, 536, 635, 1170, 5002, 5007, 5118, 5119), two have plain bows (155, 5132) and an additional one (1087) is too corroded to ascertain if there is any decoration. Mackreth (1998, 130) has suggested that the decorated ones might be considered to be early and so this could suggest a strong pre-Conquest presence. The only example of any of these from a stratified context was 630 from a Phase 1-2 pit. There are also five examples of a variant with a thin rectangular-sectioned
upper bow, a very narrow lower bow and stamped decoration on the upper
bow (142, 151, 165, 997, 5042; Fig.
5.3.5c). This is Hull Type 10D (Hattatt 1987, 22 no. 739). It
has a restricted distribution spreading from Hampshire to Gloucestershire
and Mackreth (1993, 31 no. 14) has suggested it may be an Atrebatic type.
There is also a single example with an expanded bow (154) which Olivier
(1988, 37, 15) notes as having a very restricted distribution in the south-west
(Fig.
5.3.5c). Neither of these variants is closely dated though the
presence of one of the latter at Hod Hill (Brailsford 1958, fig. 8.2)
would suggest currency during the 40s for that variant.
There are four certain examples of strip bow hinged brooches (Hattatt 1985, 68), In each case the hinge held by curving the head down and behind the bow (5004, 5023, 5028, 5135). It is possible that 5009 is another example but it is so obscured by mortar that secure identification is impossible. This is a pre-conquest type that had gone of use by c. AD 75 and the concentration of the type in the Dorset / Wiltshire / Somerset area has led Mackreth (1998, 141) to see it as typical of the Durotriges. Attention may be drawn to 5028, which is unusual in both size and catchplate design.
There is one example of the continental type with a bold angle by the head (153). This is a mid Rhine type (Riha 1979, Type 2.6) current in the first half of the 1st century with a distribution in Britain primarily south of the Thames (Crummy 1983, 8 no. 16).
Pieces of Langton Down brooches (Hattatt 1985, 35) are common in the assemblage but as all but 719 are relatively small fragments a precise number of brooches cannot be stated. There are 7 spring cover and upper bow fragments (68, 71, 433, 686, 928, 936, 5106) and three lower bow fragments (934, 1193, 5113, 5201) indicating a minimum of 9 and a maximum of 11. In all cases the bows are reeded and in most cases the junction of the bow and the spring cover is curved, with an angular stepped junction seen only on 686 and 5106. Only one example (719) was stratified. It was recovered from a Phase 1 pit. This is a continental brooch form whose use on the continent is placed principally in the first two-thirds of the 1st century (Feugère 1985, 266). Hattatt (1985, 35) suggests that Langton Down brooches were primarily a post conquest form in Britain, but the numbers recovered from the pre-Conquest cemetery at King Harry Lane, Verulamium (Stead and Rigby 1989, 91 type E), where they formed 12% of the large brooch assemblage, shows that numbers were entering the country in not inconsiderable numbers prior to the Conquest. The distribution is primarily an eastern one (Olivier 1988, 45) and so the strong presence of the type in this assemblage is of some note. None, for example, are recorded in the large assemblage at Kingscote (Mackreth 1998) where there is a strong presence of mid 1st century brooches.
There is one fragment (5087) from the style of Rosette brooch where the central disc is slightly convex and leads directly to the lower bow (Hattatt 1987, 47 Class C; Feugère 1985, Type 16a2). When complete it may have been similar to one from Colchester (Hawkes and Hull 1947, 316 no. 79). On the continent the form was in use during the Augustan period and there are grounds for believing that in Britain too it had gone out of common use by the time of the Conquest (Mackreth 1999, 219). It is likely that the lower bow and perforated catchplate fragment 5104 also came from this broad form of Rosette brooch. A second form of Rosette brooch is represented by 538 which consists of a disc and fantail without the arched bow (see Stead and Rigby 286 no. 67.4, fig. 99) Originally it is likely to have been decorated with repoussé sheet (see Feugère 1985, type 20). In Britain these appear to have been in use during the second quarter of the 1st century (Mackreth 1995, 971 no. 94). This example was found in a ditch fill of uncertain date.
Thirteen examples of one-piece Colchester brooches were found (Hattatt 1989, 24), all from effectively unstratified contexts,. This is an early 1st century form that was probably no longer being made by the time of the Conquest (Mackreth 1998, 116). Most are in very poor condition and are fragmentary (144, 145, 147, 152, 613, 1157, 1194, 5102, 5126) which makes it difficult to assign them to particular variants. 41 and 1003 seem to have the normal rounded bow section of the Colchester. The hooks on both 143 and 303 are long and this is normally regarded as a feature more typical of the later examples (ibid). In addition to the brooches that can definitely be identified as belonging to the Colchester one-piece form, there are four very fragmentary and corroded fragments (156, 631, 1136, 5112) that may either have belonged to that category, but which could equally have been the later two piece Colchester Derivative two-piece type (see below). The fragment 342 may also have been from a Colchester brooch with a strip bow. Mid to late 1st century forms
Four examples of Aesica brooches (Mackreth 1982; Hattatt 1987, 54) were recovered. One (58) is the upper fragment of a two piece construction brooch and one (1112) is fron an example with single piece construction. The other two (97, 1140) are represented by foot fragments decorated by ring and dot patterns. The only example to retain its spring fixing arrangement is the two-piece 58 which has the Polden Hill fastening. 1140 was found in a Phase 1 ditch, all the others were unstratified. It is possible that the form developed prior to the conquest (Mackreth 1998, 130 no. 70) but the main currency was during the mid to late 1st century.
There is a single example of an Eye brooch (1035). It is very corroded and so the precise variant cannot be identified, but the profile is that of Riha (1979) Type 2.7 which would place it in the middle years of the century (Claudian to mid Flavian).
There is one classic form of Aucissa present (5008) with the hinge bar held by the head rolled forward and the fragment 158 probably also came from another example. Aucissa brooches are an Augustan to Claudian form on the Continent and most in Britain arrived after AD 43. 217, however, comes from a form that is a very early variant (Riha 1979, 115 variant 5.2) for which Mackreth (1995, 974 no. 90) argues a date of c. 25 B.C to AD 10 early in the sequence. 217, therefore, may be an earlier pre-Conquest find. All of the brooches from this site are unstratified.
A single example of a Bagendon brooch was found unstratified (5110). These are a relatively rare Claudian form found scattered throughout southern England (Hattatt 1985, 53).
Hod Hill brooches (Hattatt 1985, 56) are the commonest mid 1st century form in the assemblage. These arrived in Britain at the time of the Conquest and were going out of use during the 60s. They come in a wide range of shapes and sizes but all have hinged pins, foot knobs and are frequently decorated with white metal plating. At Somerford Keynes the commonest form is the one without side-lugs where the upper bow has vertical channels and the lower bow is triangular (Hull Type 60) There are 8 certain examples (80, 83, 94, 161, 591, 1070, 5013, 5039), and the fragmentary 297 probably also belongs to this variant. Mackreth (1998, 139 no. 129) has argued that the type had gone out of use by c. AD 55-60 on the grounds of its rarity at Wroxeter. 5031 is very similar but differs in that the lower bow appears to have shallow transverse ribs rather than being a flat field on which patterns could be picked out in the white metal coating. There are two examples with side lugs placed centrally on the upper bow (Hull Type 62 - 150, 1086). There is also a particular variety of the form (1169) which probably had niello decoration on the cross bar (see Hattatt 1985, 58 no. 318), though corrosion has removed all traces of this. There are also two examples with the lugs on the base of the upper bow (Hull Type 61 - 219, 1155). The fragment 59 was probably from the variety with the lugs at the top of the upper bow (Hull Type 63). The precise form the fragments 1002, 1084 and 5123 is unclear as the edges are broken but they too can be placed in the Hull Type 60-63 range. Other forms represented include Hull Types 70 (793) and 71 (881) a type similar to a Hull Type 60 but with central button (587) and several fragmentary examples (42, 57, 76, 80, 1006, 5010) generally from the forms decorated by transverse ribs. Four of these brooches were found stratified. 793 came from a Phase 1 pit. 587, 793 and 881 came from contexts more broadly dated to Phases 2 – 3.
197 is a fragment from a mid first century disc brooch (Hull Type 239) which when complete would have had bone bosses riveted to the circular panels (see for example Hattatt 1985, 138 no. 514). It is a continental type (Riha 1979, Type 7.8) with a distribution centring on Gaul and with most British finds found in the east. The fragmentary 178 may be another mid to late 1st century form as the central perforation is found on a variety of such brooches (cf Hattatt 1989, fig. 201). Another disc brooch is represented by 5025 but this is too corroded for any typological identification to be made as to whether it belongs to the mid 1st century or later tradition.
Penannular brooches notoriously have long-lifespans but it is possible that the two examples of Fowler (1960) type D5 brooches with turned back notched terminals (1082, 5036) should be regarded as part of the mid 1st century assemblage here. The form has been found at Usk in a pre-Flavian contexts (Manning et al 1995, 94 no. 76, fig. 28), and Mackreth (2000, 157) has also drawn attention to the regular occurrence of the type in mid 1st century contexts. A third Type D penannular was also found (1166) but was too corroded for the precise variant to be identified. Mid 1st to 2nd century forms
Two-piece Colchester Derivative brooches developed shortly before or about the time of the Roman conquest and were a common mid to later 1st century form with some continuing in use into the 2nd century. There is a range of variants often with strong regional tendencies, and the spring fixing arrangements can help to date some. The simplest form is a hook facing the rear holding the spring chord. This was an ineffective method and was only used in the mid years of the 1st century. The second type has a double perforated lug behind the head with the spring chord passing through the upper lug and the bar inside the spring through the lower. Mackreth (1998, 117) has called this the Harlow method and pointed out that it is typical of the Colchester Derivatives in tribal areas of the Catuvellauni and the Iceni, and is much commoner in that area than elsewhere. The third system, termed the Polden Hill method, has the bar inside the spring lodged in perforations in the end of the wings and the chord held by a hook or lug. This is the preferred method in the west of the country. Some Colchester Derivative brooches were hinged. The basic Colchester Derivative brooch (Hattatt 1987, 88.) is represented by 13 examples. Only one (81) has the cavetto moulding typical of Hull Type 92, the rest may be placed in his Form 93 (49, 60, 146, 585, 1001, 1133, 5024, 5107, 5111, 5117, 5120, 5133; Fig. 5.3.5l: Colchester Derivative (Hull types 92-3)). It is not possible to date most of the examples from Somerford Keynes more closely within the mid 1st to 2nd century period, though on 49 and 5120 the lug has been fashioned as a skeuomorph of the hook that held the spring in the one piece Colchesters, and 146 has an elaborately perforated catch plate. Both of these features suggest that they may have be made in the mid 1st century rather than later. The only stratified example (585) came from an early 2nd century well fill. Apart from 146 where the head is too corroded to be sure of the precise spring fixing arrangements, all of the brooches have the spring held by the Harlow method. Though the eastern tendency of this method has been noted, Mackreth (1998, 117 no. 9) has drawn attention to a variant which is local Gloucestershire/Wiltshire form. This has a ridge down the upper bow which is stopped by two cross-cuts with a similar pair of grooves on the foot, and is represented here by 5024, 5117 and 5133. The lower body fragments 5092, 5124 and 5127 may also belong to this type. It is normal to consider the variant of Colchester Derivative where the head is humped over the wings as a separate type, the Dolphin (Hull Type 94; Figs 5.3.5m and 5.3.5n: Dolphin brooches (Hull type 94)). Fourteen are present in the Somerford Keynes assemblage with a variety of spring-fixing arrangements. 315, 5006 have the rear hook method of the mid 1st century. 5006 is of particular interest because the catch-plate was obviously separately made and slotted into the groove on the back of the bow. 50, 163, 317, 572, 792, 1173, 5005 and 5018 all have the Harlow method of spring fixing. On 79 and 1079 the spring is held by a forward hook and 5012 is hinged. 159 and 994 are too corroded for the precise arrangement to be ascertained, but they were obviously sprung and had some form of chord hook. 572 and 5012 have large perforations which would suggest a 1st century date, though 572 together with 792 was found in a 2nd century occupation layer. Amongst the Dolphin brooches there are four that belong to a distinctive type with elaborate flanges on the upper bow (50, 317, 792, 1173, 5018), with three (50, 317 and 792) being very similar. All have the Harlow method of pin fixing and may be a local variant linked with the Polden Hill brooches with expanded heads (Hull Type 103) discussed below. As is to be expected in this region the commonest type of Colchester derivative recovered was the Polden Hill with a total of 40 identified with certainty (Figs 5.3.5o and 5.3.5p:Polden Hill (Hull type 98)). A recurring variant has the perforated lug on the head continuing on the upper bow as a transversely notched ridge, a semi-cylindrical spring case with generally paired grooves at the ends and a perforated catch-plate. None have the semi-circular mouldings at the head/wing junction. There are five examples with a ridge (159, 326, 790, 921, 1066) and a further three where the ridge is suppressed but the line of transverse grooves is present (91, 5116, 5137). 1141 may also come from the ridged variety but is very corroded and the identification is not certain. This form has been associated with Hull Type 98 (Hattatt 1987, 97 no. 896), but Macreth in discussing six examples from Kingscote and Wycomb could not identify a sub type into which they fitted though he noted a distribution stretching from Wiltshire to North Wales (Mackreth 1998, 118 nos. 17-20). With these examples from Somerford Keynes, there is now a marked concentration in the Gloucestershire region. The scant dating evidence hitherto suggested It was in use during the second half of the 1st century (ibid). The contexts of the ones found here support this. 790 and 1141 were found in Phase 1 ditch fills whilst 326 came from a 2nd century occupation layer. Slightly different variants are represented by 664 (with a solid catch plate), 725 and 5021 (with head mouldings) and 556 with decorated wings, decorated return to the catchplate and small footknob (Fig. 5.3.5p). The decorated catch plate on 556 has been noted as a Gloucestershire trait (Cracknell 1990). Only 556 was stratified and was found in a Phase 2/3 ditch fill. Other examples of Hull Type 98 with the classic moulding on either side of the head include 85, 93, 402, 723, 1078, 1176 (Fig. 5.3.5q: Polden Hill (Hull type 98)). 402 has an elaborate perforation in the catch plate suggesting a mid to late to 1st century. 5131 is a very highly corroded Polden Hill brooch whose lower end is broken but which may have has a separately applied foot knob. This would be unusual but is occasionally found on Colchester Derivative brooches as may be seen on one from Wilcote (Mackreth 1993, 29 no. 10) from an early 2nd century context. The only stratified example was 723 from a Phase 1 gully fill. The other type of Polden Hill that is common in the assemblage is the Hull Type 103 with widened panel on the upper bow (Hattatt 1985, 84 no. 382; Fig. 5.3.5r: Polden Hill (Hull type 103)). There are three examples with a rectangular panel (96, 1146, 1158) and two where the panel is triangular (834, 1137). 1076 may also belong to this variety but is much corroded and the identification cannot be certain. This type is a south-western one and a date of c. AD 65 – 125 has been suggested (Mackreth 1993, 29 nos. 6-7). The narrow lower body fragments with small projecting foot knobs (51, 54, 65, 66, 87, 5093) are also likely to belong to this type (for similar lower bows see 834, 1137, 1146). One of these (87) has a decorated catch plate. Three brooches seem to be variants of the form. Two have narrow flanges on the upper and lower bows (201 and 1071), and one (1138) which seems to be a very strange hybrid of Hull Type 103 and the T-shaped type Form 104, another lower Severn form (Hattatt 1987, 102). Other types of Polden Hill brooch occur in far smaller quantities. 5020 has no decoration other than a pair of ribs on the wings. There is one example (1077) of the light Polden Hill (Hull Type 97). This is a mid 1st century form with a distribution stretching from south Wales through the middle Severn valley to Wiltshire. Webster (in Manning et al 1995, 74-5) has noted it is particularly popular in southern Wales. The rarity of the form in this very large assemblage might also suggest it was not made in the immediate vicinity. There is also one example of a massive embossed brooch with heavy mouldings and the chord held by a backward facing hook which should place it in the mid 1st century (1067; Fig. 5.3.5s: Polden Hill (Hull type 100)). Another brooch with heavy mouldings is 78 (Fig. 5.3.5s). This has an upper bow reminiscent of the Hull Type 103 mouldings but here combined with a central button suggesting some influence from the trumpet family. Two small brooches combine Polden Hill spring fixing arrangements with bow decoration more often seen on the T-shape brooches concentrated in the lower Severn area. 5001 (Fig. 5.3.5s) is enamelled and has a pronounced foot moulding while the fragmentary 314 and 5088 have stepped flat heads similar to Hull Type 110 (Hattatt 1987, 104 no. 10), a late 1st to mid 2nd century form. 5029 with a narrow head with Polden Hill spring fastening may also be a hybrid between the Colchester Derivative and the T-shaped tradition. Finally four lower bow fragments may be considered here. A Polden Hill brooch with similar transverse stepped mouldings to those on 89 and 576 has been recovered from Alchester (Lloyd-Morgan 2001, 225 no. 4). Two lower bow fragments have bows with a deep central channel running to the foot and large triangular perforations on their catchplates (1171, 5121). These are very similar to a Polden Hill brooch from Wilcote in a late 1st to early 2nd century context (Mackreth 1993, 29 no. 5). The pierced catch plate would suggest a mid to late 1st century date. Both 5001 and 89 have decorated return plates which seem to be a Gloucestershire trait (Cracknell 1990). Four other brooches belonging to the broad Colchester Derivative family but which cannot be assigned to any of the normal types are also present. On 1178 the spring chord would have been held by a vestigial forward facing hook but it is unclear whether the ends of the wings would have been perforated in the Polden Hill manner (Fig. 5.3.5s: Polden Hill (Hull type 100)). The side flanges are similar to those on the Polden Hill brooches 201 and 1071 whilst the moulded central block is typical of the T-shaped Hull Type 107 (Hattatt 1987, 102), and this seems to be another hybrid between the two traditions. another hybrid is represented by 52 where the spring is held by the Harlow method but the sides of the head have the mouldings typical of the Polden Hill. 218 has the spring held by a rearward facing hook and the spring axis bar would have lodged in notches in the end of the wings. this arrangement would suggest a mid 1st century date. The stepped head is reminiscent of Hull Types 109 and 110 (T shaped brooches) whilst the moulded decoration is similar to that seen on some Polden Hill brooches. Finally 43 and 880 are too corroded and fragmentary for any more precise identification . It is likely that many of the lower body fragments came from brooches of the broad Colchester Derivative family. Eight taper to narrow pointed feet, six having solid catch plates (598, 679, 1090, 1129, 5090, 5092. 5095), and two have perforations (90, 1167). Two have wider decorated bows with perforated catchplates (550, 1014). 63 also has a decorated bow but has an unperforated catchplate. Several have small ribbed foot knobs (55, 64, 77, 1132, 5125). These are most likely to come from Polden Hill brooches or the T-shaped brooches discussed below. Late 1st to mid 2nd century forms
T-shaped brooches, often hinged, are a local type popular on either side of the Severn estuary (Hattatt 1987, 100, fig. 36). Several types are represented at Somerford Keynes. There are three examples (92, 741, 796) of a form with sharply angled head, the Polden Hill method of spring fastening and an enamelled panel with lentoid mouldings on the upper body. This is Hull Type 110 (Hattatt 1987, 106 no. 910). There is also one complete (1174) and one small fragment (933) of Hull Type 122 (Hattatt 1987, 109 nos. 918-9), a hinged form with small lozenge on the centre of the bow and a small headloop. The hinged Hull Type 105 (Hattatt 1987, 102 fig. 35)with three lozenge cells down the front is represented by a single example (316), whilst 1135 seems to be a variant of Hull Type 107 (op cit). The dating evidence for both is sparse but is indicative of a floruit within the late 1st to mid 2nd century period (Hull 1967, 34-5). Unusually for this site two of the Hull Type 110 were found stratified. 796 was from a pit assigned to Phase 1 – 2 and 741 comes from a Phase 2 – 3 pit which also contained 2nd century pottery There are also two examples with ribbed bows (791, 5134). 791 has a perforated catchplate and has white metal coating suggesting a 1st century date. There are also two examples of a type with a crest and ribbed unit down the front (1172, 5022), one of which retains an acanthus foot knob. This is an uncommon form but a very similar example was found near Cirencester (Hattatt 1987, 113 no. 928) perhaps hinting at a Gloucestershire origin. There are three examples of the T-shape brooch with a serrated middle section and the Polden Hill method of spring attachment (Hattatt 1987, 116). 311 and 5109 have enamelled cells centrally, 5041 is a less common variant with incised decoration centrally. A fourth fragment, heavily obscured by mortar, is also likely to have come from this type of brooch (1131). Mackreth (1998, 123 nos. 44-7) has reviewed the scant dating evidence and suggested a later 1st century date is most likely. It is a local type with a concentration in the Somerset/ Gloucestershire area extending into Oxfordshire and Wiltshire. Finally, the foot fragment with circular cell (74) comes
from the Nor’Nour brooch (Hull Type 132 – Hull 1967, 38 Type 17) and the
5108 may be an example of Hull Type 121 Which has a distribution concentrated
in the Wiltshire / Dorset area (Hattatt 1987, 109)
Hattatt described these large ungainly brooches with good reason as ‘surely the ugliest types produced in Britain’ (Hattatt 1985, 94). An example of each type was found at Somerford Keynes (971, 1069). The dating is uncertain. An fragment from an example with a wide hinge cylinder like 1069 was recovered from a ditch fill with later 2nd to 3rd century material at Usk (Manning et al 1995, 91 no. 63), but whether this fragment was residual is unclear. Both types are concentrated in the Somerset / Gwent area with Gloucestershire being outside of its normal distribution area.
Two examples (322, 72) of the classic Backworth brooch (Hull Type 158A) with the central acanthus button running all the way around the bow. Such brooches are clearly in use by c. AD 75 (Macreth 1969, 110 no. 9) and continued in use into the mid 2nd century and possibly beyond. They have a nationwide distribution. At Somerford Keynes this classic form is out numbered by the form where the acanthus moulding occurs only on the front of the bow and there is an integral cast headloop and plate (Hull Type 158D). On these the spring is held between two pierced lugs and the central mouldings are mirrored on the back by transverse ribs. Five examples are present (95, 164, 638, 1175, 5037). The form has a distribution centred on either side of the Severn in the Gwent, Somerset and Gloucestershire area (Hattatt 1987, 134 no. 959.). Examples from well-dated contexts are relatively rare. A complete example from the Caerleon vicus was from a late 2nd century context (Lloyd-Morgan 2000, 331 no. 8). One lacking its spring and pin from Frocester was recovered from a 3rd century context (Price 2000, 39 no. 53). There may be some hints, therefore, that the form was not in use throughout the full later 1st to later 2nd century period, but belonged to the later part of it.
The commonest trumpet brooch represented is the Chester variant (Hattatt 1985, 109). It is a late 1st century form which continued in use into the early 2nd century (Mackreth 1998, 134 nos. 102-7). Ten examples (73, 75, 162, 710, 713, 770, 1080, 1159, 5017, 5030) of the classic form were found and a broken fragment (47) may be from a related type. 770 was a complete example in good condition from a 2nd century ditch, the others were unstratified. Unlike the Backworth variant which has a nationwide distribution, this is a local variant with a distribution concentrated in the southern Severn valley into the Midlands. Mackreth (1998, 134) noted that the largest concentration of Chester brooches he knew of had been found at Kingscote with eight examples. Together with the ten examples found at Somerford Keynes, this might perhaps suggest that the workshop producing the form was in the Cirencester area. There are also several fragments from other trumpet brooches of uncertain form. These include the head fragment (82), and one lower bow fragment definitely from a trumpet brooch (5114). The four lower body fragments consisting of flat-fronted bow with semi-circular ribbed foot knobs (53, 88, 301, 5089) that may also have come from trumpet brooches.
Headstud brooches (Hattatt 1987, 120) are only represented by 5115. It comes from the type where the front of the bow is transversely grooved and the grooves may sometimes be filled by enamel (Cool and Philo 1998, 30 Headstud type 3b and 6). This variant was in use by the 70s, and was commoner in the north than in the south-west.
There is one very corroded example of a hinged keyhole brooch (593). These seem to be ultimately derived from rosette brooches and a mid to later 1st century date would be most appropriate (Hattatt 1985, 178 no 632).
It is difficult to place the fragment 56 because the front of the brooch has sheared off. The headloop and the expanded head with a cast hinge casing behind suggest it might come from the fantail family, not otherwise attested in this assemblage (Hattatt 1987, 148). 2nd century forms
There are three examples of Wroxeter brooches (Hattatt 1987, 145) which have affinities to trumpet brooches but where the head is a flat plate rather than an expanded trumpet head (969, 1143, 5129). 1143 is slightly unusual as there are mouldings at the head/bow junction. As is normal with this type all the examples have very different bow decoration. Such dating evidence as there is suggests a 2nd century date (Mackreth 1995, 963 no. 27) and the context of 1143 provides useful corroboration of this as it was found in a Phase 3 ditch fill with mid to late 2nd century pottery. The type has a widespread distribution throughout the country.
There are three plate-headed trumpet brooches. 40 and 5103 have the normal acanthus moulding but 5032 has an unsual spoked button. This is a local form with a distribution centred on Gloucestershire (Hattatt 1987, 125 and Table 3). It would appear to be a mid 2nd century form as Mackreth (2000, 150 no. 21) has noted that none have been found in contexts prior to the Hadrianic period.
There was a single fragment from a very large example (98) of this sort of brooch probably originally decorated by silver foil. In general the form is poorly dated, though an unfinished example was recovered at Castleford in a context dated to c. AD 140-180 (Cool and Philo 1998, 49 no. 81).
The highly corroded small brooch 69 is of considerable interest. The combination of the trumpet head, the panel on the mid bow and the applied metal trim relate it to a pair of rare variants normally decorated with either a half disc or a pelta (Hattatt 1989, 88). The central panel now appears to be triangular but this is probably the result of extreme corrosion. Only about a dozen of either type are known and the main distribution is in the east of England. Outliers in the west country are rare (ibid. fig. 45). By analogy to the commoner full disc and trumpet brooch Olivier (1996, 257 no. 110) suggests a date of the second half of the 2nd century for the family.
There are three brooches that might fit most happily into a 2nd century milieu. None are common. The delightful 1160 imitates a salmon. It was originally coated with white metal and the cell in the eye was probably filled with enamel as was a similar one in the Hattatt collection (Hattatt 1985, 172 no. 616). Bridge brooches such as 1177 are generally considered to be a continental type but British variants are known (Hattatt 1989, 148) and the headloop on this example might suggest this is another British variant. 601 is very highly corroded but may originally have been similar to another in the Hattatt collection (1985, 158 no. 574). A fourth disc brooch is represented by 67. 4th century form
A single crossbow brooch was found (216). It belongs to Keller Type 3/4 and may be dated to the mid 4th century (Swift 2000, 15). Its presence on the site is interesting not only as providing evidence for 4th century activity, but also in hinting that the activity may have had an ‘official’ aspect as such brooches appear to have been part of the regalia of officers and administrators
The fragment 291 is from a Fowler (1960) Type E brooch when complete it would have been similar to, though smaller than, one from Aldborough (Bishop 1996, 58 no. 344, fig. 33). The form is a late Roman one, in use during the 4th century and probably into the 5th century (Mackreth 2000, 158 no. 43). Miscellaneous brooch fragments In addition to the brooches discussed above, there were four fragments from flat fronted lower bows with the stumps of catch plates centrally behind (46, 435, 1088, 5091). It is most likely that they belonged to the Rosette or Hod Hill families and thus might be of 1st century date. There are also fragments of corroded bows (61), six fragments of sprung pins (70, 84, 628, 811, 884, 5122), four fragments of hinged pins (547, 616, 918, 5105) and two fragments of catchplates (45, 48). Brooch summary With so many brooches it is felt that it might be helpful to provide a summary in Table 3. This shows the examples which can be confidently assigned to the types with the types arranged in chronological order. This makes it very clear that there is a very strong pre-conquest present which is in line with the Iron Age coinage. The mid 1st century presence is very strong too especially when it is realised that many of the mid 1st to 2nd century Colchester Derivatives etc have spring fixing arrangements that point to use at that time. The absence of the late 2nd century forms such as knee brooches is noticeable in an assemblage this size. Though they never occur in as large numbers as the earlier forms, they were being used in the vicinity. They are present, for example, at Kingscote (Mackreth 1998, 142 no. 147) and Uley (Butcher 1993, 153 no. 13). Though it is always difficult to argue from negative evidence, this might suggest that whatever the reason for such high brooch use at Somerford Keynes was, the impetus was passing in the mid 2nd century. Bracelets (Fig. 5.3.6: Bracelets)The bracelet assemblage from the site is of considerable interest despite all the examples effectively coming from unstratified contexts. In general in Roman Britain, it is the 4th century when wearing bracelets was fashionable. Often these were worn in groups rather than as a single bracelet on each arm as may be seen in numerous graves where the body was deposited wearing jewellery, see for example graves at Rochester (Cool 1981), Winchester (Clarke 1979, 67 grave 323). As a consequence fragments of 4th century bracelets tend to be represented in large quantities. Here the only examples are 228 and 991, both examples of light bangle forms of the ubiquitous light bangle forms. This scarcity probably reflects both the reduced level of occupation at the site during the 4th century and the method of collection as such fragments fall into the category that metal detecting frequently fails to find. Instead the assemblage is dominated by more massive penannular bracelets which are normally much rarer (Cool 1983, 139-44). There are three examples (249, 262, 5140) of plain bracelets with simple expanded terminals (ibid Group 5) and a further two (254, 255) where the terminal is decorated by ribs and convex units (ibid Group 7). On 250 such a terminal is combined with a grooved hoop (ibid Group 8b) and on 248 it is combined with a similar unit centrally. 1092 clearly belongs to this family but as it is represented merely by a terminal the complete decoration is unknown. The discovery of a Group 8b bracelet at Somerford Keynes is of particular interest as examples of this type show the sort of similarities to be expected of the products of the same workshop, probably based in the West country as six of the eight known in 1983 were from the Dorset / Devon / Gloucestershire / Wiltshire area (ibid 762-3). Dated comparanda for penannular bracelets such as these are rare. Plain penannular bracelets and ones with decorated terminals are an occasional find throughout the Roman period, but examples of Group 8b were clearly in use during the Antonine period. Two even rarer forms are penannular bracelets are also present. 5138 is the terminal of bracelet with twisted back snakes head terminals. These have a strongly regional distribution concentrated in the Gloucestershire area (Cool 1983, 207). None have come from a usefully dated context but the style of snakes head terminal is typical of that seen on 2nd century finger-rings (see for example Johns 1997, 105 nos. 275-9), and a 2nd century date might be suggested by analogy. 310 can be almost exactly paralleled by an example from Wroxeter (Bushe-Fox 1916, pl. XVII/23), but I know of no other bracelets like these two. The Wroxeter discovery would suggest the piece is of Roman date but no closer dating than that can be advanced. Finally 5142 amongst the copper alloy bracelets there is an example of a wide cuff bracelet (Cool 1983, Group 9), a type that was current during the mid to late 1st century with a distribution centred on the early civilian centres of London, Verulamium and Colchester (see for example Crummy 1983, 36 nos. 1586-7; Waugh and Goodburn 1972, 120 nos. 30-31). It has been noted that when they occur outside of this area the sites are often associated with evidence of military activity (Cool 1983, 146). There is also one fragment of a glass bangle (868). How such items were used is the matter of some debate and they may not all have been arm ornaments in the way most copper alloy ones were (Price 1988, 354). 868 has a large diameter and so could have been a bracelet and will be discussed here. It is an example of Kilbride-Jones (1937-38) Type 2. Large examples such as this appear to be predominantly a south-western variant in use from the Claudian to early Neronian period (Manning et al 1995, 100-102). Finger rings (Fig. 5.3.7: Finger rings)Seven finger rings of the typical 1st to 3rd century form (Henig 1974, Type II) were recovered, all from unstratified contexts. 214 retains the remnants of what appears to be a moulded green glass intaglio. This dates the ring to the 3rd century (ibid 164). 588 retains decayed enamel in the bezel while 207, 1068, and 5100 all have deep blue glass settings. The bezel of 594 is very corroded and any decoration it may have had is now obscured. In 212 the bezel setting is empty. With the exception of 214, here the bezel setting is present it is always decorative rather than a utilitarian intaglio. This would suggest that the rings were most likely to have been used in the 2nd to 3rd century than earlier. In the assemblage there are a further three finger rings with expanded bezels of broadly similar form (204, 209, 213), the Roman date of these is much less certain and it is probable that they are of more recent date. There is a single example of an enamelled ring with notched shoulders (202). These are a relatively common 2nd to 3rd century form (See Cool 1983, 251 Type 13A). 1085 might also belong to this broad type but is too corroded for the identification to be certain. Another very corroded finger ring (1074) gas a cable-twisted hoop and is likely to have been very similar to two silver rings from Caerwent and Silchester (Cool 2000, 31 footnotes 38 and 39) and one from the Snettisham hoard (Johns 1997, 110) which suggest the type was in use during the mid 2nd century. BeadsOf the three beads that were recovered, only the annular blue/green glass bead 280 is a common form. They are a 1st to 2nd century type but are occasionally found later (Guido 1978, 65 Group 6iib). The other glass bead (781) is puzzling as annular beads made in glass that appears black are rare in the Roman period (Guido 1978, 68) but not uncommon in 5th to 6th century contexts (Guido 1999, 20). Given there is very late Roman material in the small find assemblage which probably indicates occupation into the 5th century, there is a possibility that a 5th century bead could be present on the site. 781, however, came from a Phase 1 context. there is also a fragment of a large globular bead from a Phase 2/3 context. Such beads are very uncommon finds. HairpinThe only example of a hair-pin in the entire assemblage is 298. This is an example of a form that is common in the south-west during the 2nd century (Cool 1981, 164 Type 13). The fact that only a single hairpin was found is probably an example of the bias in metal detected assemblages which don’t appear to locate long thin artefacts with any ease. The virtual absence of worked bone amongst the finds from Somerford Keynes also contributes to scarcity of this artefact type as in most assemblages the, presumably cheaper, bone hair pins were much commoner than copper alloy ones. Here their presence is only hinted at by the shank fragment 773 from a Phase 3 context which could as easily come from a needle as a hairpin. Shoe cleatsTwo shoe cleats were recovered, 761 and 817 from phase 2-3 contexts. The absence of hobnails is probably due to the deterioration of the iron since excavation. They are normally recognised from X-radiography but the years in storage prior to the X-radiography here would most likely have reduced them to undiagnostic fragments or ‘nail’ heads. Buckles Within a Roman context buckles are normally a military fitting and Roman buckles will be considered below. One buckle that may be noted here is 194 which is of 14th century date (Hinton 1990, 507). The belt plate 1057 is also more likely to be of medieval date than Roman. Toilet EquipmentThe toilet equipment can be divided into two broad categories. By far the commonest are small implements for personal use, but there are a few examples of long-handled implements that would have had a role both in personal care and in medicine. In addition there is a single fragment of a mirror (1091) from an unstratified context. This seems most likely to have come from a mirror of Lloyd-Morgan Type X (Lloyd Morgan 1981, 91) which are considered to be a 3rd century type. Small implements (Figs 5.3.8a and 5.3.8b: Toilet Equipment)Most of these clearly came from toilet sets where a group of implements were held together by a fastener. The only implements found together and clearly indicating that they were part of a set were 5026 and 5027 consisting of a pair of tweezers and a nail cleaner (Fig. 5.3.8a: Toilet Equipment). Tweezers tend not to be chronologically sensitive but the shape of the nail cleaner is similar to a type found in the 1st century (Crummy 2001, 3 fig. 1b - see for example Hands 1998, 60 no. 79, fig. 21) though this example lacks the central groove and has edge grooves instead. Another toilet set is probably indicated by 574 which is likely to have been the handle of a set of toilet implements which would have been threaded onto the crossbar inserted into the two perforated terminals as may be seen on an example from Colchester (Crummy 1983, 62 no. 1943, fig. 67). Four other nail cleaners were found as individual implements but have perforated terminals indicating they were part of sets. Objects 210 and 709 (Fig. 5.3.8a) belong to the same family as the one in the toilet set. Object 586 (Fig. 5.3.8a) is another very simple sheet implement, a similar one was found in a context of the 1st half of the 2nd century (Hands 1993, 38 no. 24). It is also possible that a late 4th century nail cleaner may be represented by 693 (Fig. 5.3.8b). In shape this is very reminiscent of late 4th century styles of nail cleaner but those normally have decorated faces, see for example one from Gestingthorpe (Draper 1985, 36 no. 77; see also Crummy 2001, 6 fig. 5). The surfaces of this one are pitted by corrosion but there is no evidence of any decoration other than the groove. There is one example of the nail cleaner form which had a bone terminal (1110), see for example one from a mid 2nd century context at Wilcote (Hands 1998, 60 no 77, fig. 21). Crummy notes that this is a south-western type which comes into use in the mid 2nd century and possibly continues into the 4th century (Crummy 2001, 4). The only certain cosmetic spoon from a toilet set is 894 from a Phase 1 context. This is a very simple form not closely dateable other than by its context. There is also one example of an implement that could either have been from a cosmetic spoon or a nail cleaner (1094; (Fig. 5.3.8b). Cast nail cleaners with similar cross-hatched cylinders have been recovered from mid 2nd century contexts at Wilcote (Hands 1993. 38 no. 22; 1998, 60 no. 78) and again this seems to be a south-western type (Crummy 2001, 4 fig. 3b). Tweezers are remarkably common in this assemblage with a minimum of 18 certainly present judged from the complete or fragmentary ones that retain the upper loop. In addition there are 10 fragmentary arms. Allowing for the fragmentary state of some in the first category, these arms must represent at least three others. Taken with the example in the toilet set, the entire assemblage therefore consists of a minimum of 22 or a maximum of 28. The commonest form is a simple undecorated strip bent into shape with a closed loop (205, 220, 730, 737, 799, 878, 955, 1151, 5073, 5074, 5075. Objects 294 (Fig. 5.3.8b) and 5071 are similar but have diagonal grooved decoration on the arms. Open loops are less common but tend to be decorated. On 607 the decoration is on the loop whilst on 223 and 5072 there is grooved decoration on the arms. Where the shape of the loop is uncertain or only arms are present, the majority are plain (196, 221, 222, 224, 226, 227, 229, 584, 1096). 225 has grooved decoration and on 5076 there are grooves parallel to each edge as on the set of tweezers in the toilet set. Tweezers such as this rely on their contexts if they are to be dated closely. Here nearly all of them were found unstratified though 799, 730 and 878 came from contexts of Phase 1, Phase 2-3 and Phase 3 respectively. There is also one fragment from a set of tweezers of different construction (267) where a bar has been split and the tweezers are provided with a collar. It would have been possible to take a more precise grip with such an implement and it may have had a more specialised purpose than the tweezers discussed above but in the absence of the jaws it is not possible to explore this possibility further. The number of tweezers in the assemblage does seem exceptional as may be seen from Table 4 where the numbers of tweezers and nail cleaners at various local sites are considered. It is possible that the large percentage from Somerford Keynes is a result of the under recovery of nail cleaners. If the examples published from Wanborough (Hooley 2001, fig. 44) can be taken as typical of the area, which seems likely, it is noticeable that 71% are long and thin and so might be seriously under-represented in a metal-detected assemblage. Long-handled implementsThese are much less common than the small implements. 86 and 208 are fragments from long-handled cosmetic spoons whilst 269 and 846 are olivary probes probably from double-ended implements. Like the nail cleaners, these belong to a shape category that can be expected to be under-represented in a metal-detected assemblage. All four of those found were unstratified and are types which are not chronologically sensitive within the Roman period. Textile EquipmentThe only items that can be assigned to this category are two spindle whorls and a possible loom weight. The spindle whorls made from re-used pottery sherds, 782 from a Phase 3 context and 5045 was unstratified but has the typical narrow spindle hole typical of one of Roman date. 869 from a Phase 2a context may be a fragment of a triangular loom weight an Iron Age to early Roman form (Elsdon and Barford 1996) Household utensils (Fig. 5.3.9: Household utensils )Six items of household equipment made of metal were found unstratified. Most can be dated to the Roman period with some certainty. Two are definitely of early Roman date. 629 is a broken fragment of a round-bowled spoon . Such spoons (Crummy 1983, 69 type 1) are a common 1st to 2nd century form. By contrast the other item (198) is a much less common find. It is a flattened tankard handle (198) of Corcoran (1952) Class V, a type found in 1st to 2nd century contexts. An example from Newstead, for example, was recovered from a pit filled during the late 70s - 80s (MacGregor 1976, volume 2 no. 290). Tankards appear to have been a popular vessel form in the south-west, not only are wooden examples with metal fittings known but pottery examples are more regularly part of the ceramic repertoire in the region than elsewhere, for example in Severn Valley Wares (Tyers 1996, 197). 1055 may also belong to this group of early material. It seems most likely to be the foot of a patera or bowl, certainly the differential treatment of the front and back and the shape of the piece is consistent with this interpretation. It has to be noted, however, that piece is slightly smaller than would be expected for one of these feet, and the presence of the ring and dot decoration would be unusual. Pelta feet were applied to the underside of the base to raise it slightly above the level of the surface in which it was placed. They were used on the bowls belonging to the Hagenow style of jug and patera sets of the early to mid 1st century (Nuber 1972, 38) such as that from Snailwell (Lethbridge 1953, Pl VII). They are also known on other styles of bowls (see for example Tassinari 1993, 241). The second spoon (346) from the site is of late Roman date. Spoons very similar to this have been found at Chignall was found in a context of AD 285-370 (Major 1998, 79 nos. CA69-70) and another came from Barnesley Park in a 5th century context (Webster and Smith, 1982, 110 no. 134). 346 not been scientifically analysed but the small points of green corrosion products visible here and there suggest it is of copper alloy which has been coated with white metal (silver or tin) to produce the effect of being made of silver. The other items cannot be assigned to the Roman period with any certainty, though in the virtual absence of medieval material in the assemblage is it very difficult to argue for a later date. There is one virtually complete copper alloy dish (5197). This could be of Roman date but unfortunately the vessel is of such a simple form that it is impossible to date typologically. The jug handle 141 certainly does not belong to any of the well-known Roman jug handle types. Weighing Equipment (Fig. 5.3.10: Weighing Equipment )There are four items that can be interpreted as steelyard weights. 951 is a biconal lead weight with iron suspension loop. This is the typical Roman form as can be seen on a steelyard from the Walbrook. London which retains its weights (Merrifield 1965, pl. 128). The hemispherical lead weight (279) and the cylindrical weight (614) both have iron loops passing through them in a similar manner, and though they are not typical steelyard weight shapes, they may have functioned as such. There is also one copper alloy weight in the shape of an acorn (5040). In Britain items such as this are normally interpreted as steelyard weights, which the large ones such as that from Castleford in a late 1st century context (Cool and Philo 1998, 94 no. 445, fig. 35), almost certainly must be. Oldenstein (1977, 159) has argued for the smaller ones being amulets, although small ones too can have the sort of inserted iron loops typical of a range of steelyard weight types as can be seen on one from a 4th century context at Caerleon (Webster 1992, 157 no. 366). In addition to these 947 may be noted. It is very similar to a typical steelyard weight like 951 but has what appear to be additional wire loops passing through the centre. Whether it too served as a steelyard weight is uncertain. There is also a single weight for an equal-armed balance (1045). It belongs to the common cheese-shaped form and retains two dots. Such a marking would indicate that it was intended to be a weight for a sextans (2 unciae). Such a weight should weigh either 54.58gm or 54.25gm (RIB II.2, 2) and so the extant weight of the piece (54gm) would be appropriate allowing for the inevitable weight loss through oxidation. Writing Equipment (Fig. 5.3.11: Writing Equipment )Five items may be assigned to this category. There are two styli and unusually for a Roman site neither are made from iron. Given the problems with the survival of iron on this site, such an absence is probably not significant. One (1145) is made of copper alloy and was recovered from a Phase 2-3 context. Copper alloy styli are generally less common than iron examples. It may be noted, however, that Wilcote has also produced two with simple grooved decoration on the eraser as seen on 1145 from 2nd century contexts (Hands 1993, 38 no. 16; 1998, 58 nos. 57), and it may be that this is a local type. The other (803), also from a Phase 2 -3 context, is made of bone. It is a rather roughly fashioned item but does retain both the point and the blunt eraser that would be needed. There are also two seal boxes, both from unstratified contexts. One is a leaf-shaped box (139) for which a 2nd century date can be suggested as it is identical to an example from Castleford found in a mid 2nd century context (Cool & Philo 1998, 101 nos. 497). The other (270) was circular but is only represented by the base and so a closer date within the general 1st to 3rd century currency of seal boxes cannot be suggested. The identification of the final item (1106) as an item of writing equipment is advance more cautiously. It is a flaring iron blade with a narrow tang. The narrowness of the tang suggests that the handle was not expected to have to deal with a high degree of force, as might have been expected if it were to have been used as a carpentry tool for example. It can be suggested that it might have been a wax spatula used in the preparation of writing tablets. Such items are being increasingly recognised amongst the material culture from Roman Britain (see for example Crummy 203; Boon 1991 fig 4g & k). The tang would indicate that the implement had a wooden or bone handle. I have found no comparanda for spatulas with such handles but the it might be expected that as more are recognised, some may be found as there were clearly a variety of ways of providing these items with handles. The types of writing equipment found at this site may provide some clues as to its status. It has been shown that styli are regularly found on sites that go someway down the settlement hierarchy whereas seal boxes appear to be less common in the countryside (Cool and Baxter 2002, 375-6). It has been suggested that this might imply that though literacy was not uncommon, the sorts of documents that needed seal boxes were more an urban than rural phenomenon. The presence of seal boxes at Somerford Keynes, therefore, that this is something more than a modest rural site. TransportThere are two undoubted examples of Roman transport equipment. 167 certainly and 1049 most probably are knobbed terrets (MacGregor 1976, 46 and Map 10). Though most of the form are found in the north there is a distinct cluster around the Severn Valley into which these examples fall. There is some evidence they were in use in the later 1st century but most are of 2nd century date. There is also a possible strap junction (683). A Roman date cannot be advanced with certainty for this piece as it may be a relatively modern horse-trapping. It may be noted, however, that similar, though smaller, trappings are found on Roman sites (Allason-Jones 1988, 183 no. 210). Buildings and ServicesThe only items considered in this section are those that were found securely stratified. They are summarised in Table 5 by Phase. As is normal nails predominate. The L clamp (789) and the joiners dog fragment (837) would have both been used in structural carpentry. It is also possible that a drop hinge fragment is present (819). It should be noted that it is probable that originally more structural iron fittings were excavated but the deterioration of the ironwork in the years since excavation have reduced them to featureless bars strips and fragments. In addition to these items there is also a large fragment of fired clay (904) from a Phase 1 context that might have come from a thatch weight. Tools (Fig.
5.3.12: Tools
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||