² The excavation was mainly funded by the Manpower Services Commission. Support from that organisation terminated in late 1988, prior to actual post-excavation work commencing. Cadw: Welsh Historic Monuments, though not the main funding body for the excavation, agreed to make money available, within the overall programme for rescue archaeology in Wales, to assist the Trust to meet their post-excavation commitment. This was part of a three-year funding programme designed to deal with the Trust's post-excavation commitment mainly arising from fieldwork projects which had received MSC and Cadw funding. The Trust expressed reservations about the overall level of funding being made available by Cadw for this work.

Constraints both of funding and of available staff-time within the Trust, due to the need to complete other elements of this post-excavation programme, meant that the Trust and Cadw had to agree that the report should take its present form. This is essentially a digest description of the principal structures and features encountered, accompanied by a discussion and finds report (excepting the report on the sediments and intaglio) to provide a basic chronology for the activity recorded. Other categories of artefacts have not been reported on or examined, apart from an initial screening of copper-alloy and ferrous objects in advance of their conservation at University College, Cardiff under the Cadw: Welsh Historic Monuments contract.

This approach is considered by the author to be inadequate, but Cadw were unable to provide additional financial support from within the overall budget for rescue archaeology in Wales, despite representations from the Trust. The artefacts, including the outstanding categories of finds, have been deposited in the Legionary Museum at Caerleon and may be examined on application to the curator. It is hoped that it may prove possible to report on some aspects of this material in the future.

³ Ptol. Geog. II.3.12; Ant. Iter. 4845 (Iter XII), 485₁ (Iter XIII); A.L.F. Rivet and C. Smith, The Place-names of Roman Britain (1979), 285.

⁴ Camden, Britannia (1610), 636. For a summary of antiquarian observations on Usk see W.H. Manning, Report on the Excavations at Usk 1965–1976. The Fortress Excavations 1968–1971 (1981), 3–7.

⁵ Boon, G.C., ‘Remarks on Roman Usk’, Monmouthshire Antiq. I.ii (1962), 28–33.Google Scholar

⁶ For a summary of early excavation work at Usk see Manning, op. cit. (note 4), 7–12.

⁷ Manning, op. cit. (note 4) and idem, Report on the Excavations at Usk 1965-1976. The Fortress Excavations 1972–1974 and Minor Excavations on the Fortress and Flavian Fort (1989).

⁸ Evans, D.R. and Metcalf, V.M., ‘Excavations at 10 Old Market Street, Usk’, Britannia xx (1989), 23–68.CrossRefGoogle Scholar

⁹ op. cit. (note 4), 65, fig. 1; op. cit. (note 7), 156–7, fig. 70.

¹⁰ As may have been the case with the east fortress ditch, Manning, op. cit. (note 4), 66; a column sample was taken throughout the ditch fill in Area C, but no funds have been available for analysis.

¹¹ op. cit. (note 4), 66.

¹² These were provisionally identified during the observation of a sewer-pipe construction in 1970 and confirmed by excavation in 1973: Manning, op. cit. (note 4), 65, 101; op. cit. (note 7), 156, figs 3 and 70.

¹³ Manning, op. cit. (note 4), 86–7, figs 25–6.

¹⁴ The effective limit of a Roman javelin was c. 30 m. G. Webster, The Roman Imperial Army (1969), 173 and fig. 29, suggests 40 yards; A. Johnson, Roman Forts (1983), 49 and fig. 27, suggests 25–30 m.

¹⁵ Manning, op. cit. (note 4), 119.

¹⁶ ibid., 121; Manning, op. cit (note 7), 9.

¹⁷ Manning, op. cit. (note 4), 129.

¹⁸ ibid., 149–53, 163–5, 190–5, figs 56–9, 63d, 64, 65, 81c, 82–4; Manning, op. cit. (note 7), 25, 32, 37, 48, 50, 56, 59–60.

¹⁹ op. cit. (note 7), I49–53. figs 56–9.

²⁰ ibid., 163–5, figs 64–5.

²¹ ibid., 164–5, fig. 65.

²² Manning, op. cit. (note 4), 105–6, fig. 35; op. cit. (note 7), 156, fig. 70.

²³ Manning, op. cit. (note 4), 40–54, 102–10.

²⁴ Manning, op. cit. (notes 4 and 7), passim.

²⁵ It is extremely unlikely that there was no veranda. Although some barracks of Augustan/Tiberian date do not possess verandas, their absence is extremely rare in barracks of later date. Davison notes only two certain examples: D.P. Davison, The Barracks of the Roman Army from the 1st to 3rd Centuries A.D., BAR Inter. Ser. 472 (i–iii) (1989), 18–20. It is possible that the veranda was free-standing, and may have been removed when the building was modified in Period II. Alternatively the posts rested on the edge of Street 182. If this were the case then the full width of the building would have been c. 9.8 m.

²⁶ Dormagen: G. Muller, ‘Ausgrabungen in Dormagen 1963–1977’, Rheinische Ausgrabungen XX; Krefeld-Gellep: Reichmann in litt.; see Discussion below.

²⁷ See Report on Sediments by J. Crowther below.

²⁸ ibid.

²⁹ Davison, op. cit. (note 25), fig. 20.

³⁰ op. cit. (note 7), 173–81, figs 73–4.

³¹ As V.G. Swan, The Pottery Kilns of Roman Britain, RCHM(E) Sup. Ser. 5 (1984), 30, fig. II.iv and v.

³² Manning, op. cit. (note 7), 162–70; Metcalf and Evans, op. cit. (note 8). In the latera praetorii, as defined by Manning, op. cit. (note 7), 170 n.2, the position of three granaries, two fabricae, and one of the senior officer's houses has been established by excavation and the position of the principia, praetorium, and other officer's houses can reasonably be inferred. A further series of granaries has been identified in the praetentura sinistra. In the retentura dextra the position of the baths was confirmed in 1965, and a further granary and a possible barracks were uncovered in 1979.

³³ op. cit. (note 7), esp. figs 71–2.

³⁴ As the author has noted elsewhere, ‘Usk’, Trivium xxv (1990), 23Google Scholar, Usk is virtually the same size as Colchester (an earlier base for legio XX). Crummy, P. (Colchester Archaeological Reports 3 (1984), 3–9Google Scholar, fig. 6 and pers. comm.) has argued that, at the latter site, the fortress was initially divided into strips 200 pM or 300 pM wide (Crummy's measurements exclude the intervallum road). At Usk, if we assume that the distance between the west defences and the inner edge of the via sagularis was the same as on the northern side of the fortress, then the width of the eastern plot in the northern scamnum (i.e. that defined by the intervallum street, and Streets 1626 and 1723 on the Old Market Street ‘;86’ site) is 89 m (301 pM) and the depth 62 m (210 pM) if Street 1723 formed the southern side of the plot. An alternative is that the plot was carried as far as the street (2085) separating Buildings 5 and 6. This street appears to carry on the line of an east-west road represented only by two parallel gullies found on the 10 Old Market Street site in 1979 (Evans and Metcalf, op. cit. (note 8), 29) and would give a depth for the scamnum of 80 m (270 pM). At the contemporary fortress for legio II Augusta at Exeter, the cohort blocks, including the intervallum road, measure 300 pM by 240 pM, C.G. Henderson ‘Aspects of the Planning of the Neronian Fortress of legio II Augusta at Exeter’, in M.J. Dobson and V.A. Maxfield (eds), Roman Frontier Studies 1989 (1989), 73–84. The metrication at Usk appears to be broadly similar to both sites but clearly needs further consideration.

³⁵ See note 25 for a possible explanation of the apparent absence of the veranda, which if correct would give an overall width of c. 9.8 m.

³⁶ L.F. Pitts and J.K. St Joseph, Inchtuthil. The Roman Legionary Fortress, Britannia Monograph 6 (1985), 173–7, tables VA, VB, and VIII; Davison, op. cit. (note 25), 21–40, figs 13, 16, 20 and 21, tables V and VI.

³⁷ An arrangement, perhaps not surprisingly, similar to that in the north-west quarter at Inchtuthil; Pitts and St Joseph, op. cit. (note 36), fig. 84. For a wider discussion of the arrangement of cohort-blocks with regard to the layout of legionary fortresses see Davison, op. cit. (note 25), following H. von Petrikovits, Die Innenbauten römischer Legionslager während der Prinzipatszeit (1975), 108ff. Without the information for the layout in the other scamna at Usk, it would be unwise to speculate which of the possible schemes of layout for the arrangements of the cohort-blocks around the fortress (von Petrikovits' Types 1 and 2, Davison's Types A and B) was preferred, or which of the cohorts were to be quartered in this part of the fortress.

³⁸ For summaries see, Davison, op. cit. (note 25), ch. 5; K.R. Dixon and P. Southern, The Roman Cavalry from the First to Third Century A.D. (1992); Holder, P., The Auxilia from Augustus to Trajan, BAR Int. Ser. 70 (1980)Google Scholar; Bishop, M.C., ‘Cavalry Equipment of the Roman Army in the First Century A.D.’, in Coulston, J.C. (ed.), Military Equipment and the Identity of Roman Soldiers, BAR Int. Ser. 394 (1988).Google Scholar

³⁹ The types are summarised by Davison, op. cit. (note 25), 148–60, figs 96–100.

⁴⁰ C.M. Wells, ‘Where did they put the horses? Cavalry stables in the early Empire’, in J. Fitz (ed.), Limes: Akten des XI Internationalen Limeskongress (1977), 659–65.

⁴¹ Muller, op. cit. (note 26).

⁴² Davison, op. cit. (note 25), Type J; morphologically the buildings at Dormagen are more simliar to Building 3 than Building 5 at Usk.

⁴³ Normally one per room but some rooms contain pairs of pits.

⁴⁴ Schröder in Muller, op. cit. (note 26), 129.

⁴⁵ Knörzer in Muller, op. cit. (note 26), 131–4.

⁴⁶ Muller, op. cit. (note 26), Areas B and C, see Taf. 8.

⁴⁷ ibid., Areas D and C; the example in Area C was subsequently modified so that both pairs of rooms contained pits as shown on Taf. 8.

⁴⁸ ibid.

⁴⁹ C. Reichmann in litt. Periods II–IV, but particularly Periods II–III.

⁵⁰ ibid.; for a detailed discussion of the Krefeld-Gellep buildings see Davison, op. cit. (note 25), 155–8.

⁵¹ The three more complete examples are in form Davison's Type J, op. cit. (note 25).

⁵² These are normally aligned transversely across the building, most rooms contain a single example, but there are a few rooms where the pits are aligned with the long axis and also where there are pairs of pits. In the two easternmost buildings there are two examples where there are three divisions across the buildings with two of these containing pits.

⁵³ Davison, op. cit. (note 25), 158.

⁵⁴ C. Reichmann in litt.

⁵⁵ ibid.

⁵⁶ ibid. Three of the Buildings (1, 2, 3 ) are Davison's Type J, the other (Building 4 ) is probably Davison's Type H (?C), Davison, op. cit. (note 25), 155, 165–8. Including verandas, Buildings 1–3 are c. 12 m wide, Building 4 c. 9 m wide

⁵⁷ C. Reichmann in litt.

⁵⁸ Dixon and Southern, op. cit. (note 38), 193, view the Dormagen evidence as ‘irrefutable’ and Davison, op. cit. (note 25), 151, notes that ‘the findings from Dormagen and Krefeld-Gellep …… show that there is nothing inherently impossible in the accommodation of men and horses in the same building’. On the other hand Dr Reichmann informs me that H. Schönberger, D. Baatz, and H. von Petrikovits do not at present accept the interpretation.

⁵⁹ op. cit. (note 25), 143.

⁶⁰ We can reasonably assume that stables were cleaned out on a daily basis: Xenephon, De Re Equestri v.2; Pollux, Onomasticon 1.183 (Davison, op. cit. (note 25), 139).

⁶¹ Arrian, Ars Tactica, passim.

⁶² Schröder, op. cit. (note 44); Knörzer, op. cit. (note 45).

⁶³ Crowther below.

⁶⁴ Muller, op. cit. (note 26).

⁶⁵ A process which in itself would have caused some generation of heat, and thereby provided additional warmth to the stables.

⁶⁶ Crowther below.

⁶⁷ ibid. The lack of phosphate enrichment may have occurred because the pits had a finite storage capacity and would have had to be cleaned out regularly, as is clearly evident.

⁶⁸ For exercises, patrols, grazing etc.

⁶⁹ Thus latrines are normally sited close to defences; at Usk the predominant form of latrine appears to be a deeply cut soakaway; see above and notes 18–21.

⁷⁰ Crowther below.

⁷¹ M.H. Hayes, Stable Management and Exercise (6th edn, A Horsburgh-Porter), 191–6.

⁷² Columella, De Re Rustica VI.30.2; Xenophon, Cyropaedia IV.3–5, recommends stone as the best arrangement. Dixon and Southern, op. cit. (note 38), 201, note in addition to straw the use of peat, moss, and sawdust as bedding mediums; see K.D. White, Roman Farming (1970), 290, for a discussion of the use of boards.

⁷³ von Petrikovits, op. cit. (note 37); Pitts and St Joseph, op. cit. (note 36); see also H. v. Petrikovits, ‘Militärische Fabricae der Römer’, in D.M. Pippidi (ed.), Actes du IX Congrès International D'Études sur les Frontieres Romaines (1974), 399–407.

⁷⁴ Chester, Deanery Field Block B : crucibles, vitreous slag with traces of bronze, bronze waste, and burnt clay came from the arma, waste bronze and slag and another crucible were also found in a papilio in the same building, R. Newstead, ‘Report on the excavations on the site of the Roman fortress at the Deanery Field Chester’, Liverp. Ann. Arch. Anth. xv (1928), 11. Nanstallon, Barrack IV: here the excavators suggest the working of precious metals in the end contuburnium, A. Fox and W. Ravenhill, ‘The Roman fort at Nanstallon, Cornwall’, Britannia iii (1972), 83. Pen Llystyn Barracks R2, R3, R6: these contained evidence of iron-working, A.H.A. Hogg, ‘Pen Llystyn; a Roman fort and other remains’, Arch. Journ. cxxv (1968), 138; Ravenglass Barrack 9: here the presence of slag is believed to have implied smithying activity, T.W. Potter, Romans in North-West England, Cumbd Westd Antiq. Arch. Soc. Res. Ser. I (1979), 44. Strageath Antonine I Building XVII: this has three ‘pear-shaped’ furnaces for metal-working in the end contuburnium, S.S. Frere and J.J. Wilkes, Strageath: Excavations within the Roman Fort 1973–86, Britannia Monograph 9 (1990), 112. Carzield: the officer's quarters have a furnace for metal-working and possible broken objects as scrap material, Birley, E.B. and Richmond, I.A., ‘The Roman fort at Carzield’, Trans. Dumfr. Gall. Nat. Hist. Arch. Soc. xxii (1938–1940), 161.Google Scholar

⁷⁵ Metal-working has also been identified in Room 5 of one of the officer's houses at Caerleon, Zienkiewicz, J.D., ‘Excavations in the Scamnum Tribunorum at Caerleon: the Legionary Museum site 1983–5’, Britannia xxiv (1993), 27–140CrossRefGoogle Scholar, esp. 54–5, where small furnace-bases, trays to collect metal filings, and extensive deposits of ash and hammer-scale are present. Zienkiewicz, pp. 75–6, identifies this building as the house of the praefectus castrorum. One of the tribune's houses at Colchester contained crucibles for smelting brass, Britannia xiv (1983), 309–10Google Scholar and fig. 15, but here there is a possibility that houses were still standing after the foundation of the colonia and that the activity is civilian – a possibility overlooked by Zienkiewicz, p. 57 n. 32. Parts of a craftsman's (perhaps an armourer's) tool-box, and more importantly its contents were recovered from a praetorium (Building 3.10) at Loughor, M. Dawson, ‘The Ironwork’ in A.G. Marvell and H.S. Owen-John, Leucarum: Excavations at the Roman Auxiliary Fort at Loughor, West Glamorgan 1982–4 and 1987–8, Britannia Monograph 12 (forthcoming). See also M.C. Bishop, ‘The military fabrica and the production of arms in the early Principate’, in M.C. Bishop (ed.), The Production and Distribution of Roman Military Equipment, BAR Int. Ser. 275 (1985), 1–42.

⁷⁶ Krefeld-Gellep Period IVb, Davison, op. cit. (note 25), 155.

⁷⁷ ibid., 133, and tables B and C for a summary of classical sources; see also note 38.

⁷⁸ R.M. Luff, A Zooarchaeological Study of the Roman North-Western Provinces, BAR Int. Ser. 137 (1982). See Davison, op. cit. (note 25), 145–6, for a summary of the evidence from, in particular, Corbridge, Newstead, Krefeld-Gellep, and Saalburg.

⁷⁹ ibid.

⁸⁰ For a discussion of this see Davison, op. cit. (note 25), 145–8, and Bishop, op. cit. (note 38).

⁸¹ See below.

⁸² in litt.

⁸³ The horses were presumably tethered so that their feet avoided the stable pits. In the cases where the pits run across the centre of the stall, they were presumably tethered so that their front and rear feet stood either side of the pits. These probable arrangements in themselves do not, however, provide any evidence to support the stabling of stallions in one type of stall and mares in another. See also note 89.

⁸⁴ At Krefeld-Gellep 0.50–0.80 m.

⁸⁵ Davison, op. cit. (note 25), fig. A, Types O, R, S and T.

⁸⁶ ibid., Types G and H.

⁸⁷ e.g. Ilkley, Hartley, B.R., ‘The Roman fort at Ilkley: excavations of 1962’, Leeds Phil. Lit. Soc. xii (1966), 23–72Google Scholar, esp. 36–7; Maryport, , Jarrett, M.J., Maryport, Cumbria: a Roman.Fort and it's Garrison, Cumbd Westd Antiq. Arch. Soc. sup. ser. 22 (1976), 38–9Google Scholar; Brough-on-Noe, , Jones, G.D.B. and Wild, J.P., ‘Excavations at Brough-on-Noe 1968’, Derby. Arch. Journ. lxxxvi (1968), 89–93Google Scholar, esp- 91–2.

⁸⁸ Davison, op. cit. (note 25), 151–6 and table F. In the case of some buildings double, or even quadruple, rows of horses tethered either side of a central passage are envisaged, as, for example, at Longthorpe Building vii, Davison, 151, after Frere, S.S. and Joseph, J.K. St, ‘The Roman fortress at Longthorpe’, Britannia v (1974), 1–129CrossRefGoogle Scholar, esp. 26–7; Ilkley, Davison, 152, after Jones and Wild, op. cit. (note 88); Gelligaer, Davison, 152; Haltonchesters, Davison, 153, after Simpson, F.G. and Richmond, I.A., ‘The fort on Hadrian's Wall at Halton’, Arch. Ael.⁴ xiv (1937), 151–71Google Scholar, esp. 165–7; Niederbieber, Davison, 153; Saalburg, Davison, 153; Hod Hill, Davison, 154, after I.A. Richmond, Hod Hill II: Excavations carried out between 1951 and 1958 (1968), 82–4, see also Frere and St Joseph, loc. cit., 27, n. 48. A problem with these calculations is that they are derived from attempts to fit one or more turmae into the overall space: where little evidence has survived for the form of the internal arrangements of these buildings, the horse-width allowances must be treated with caution. Dixon and Southern, op. cit. (note 38), 182–3 note tnat modern practice allows a stall measuring 3 m by 1.8 m for a hunter with a space of 2.4 m behind or 4.8 m in the case of a double row of horses; each horse is, therefore, allowed an area of c. 5.5 m². They note that this accords well with Dormagen, where 2/3 horses were accommodated in stalls that allowed a space of c. 4.5–6.7 m² per horse, but that stall allowances at some other sites are much smaller for the numbers of horses envisaged: Benwell 4.7 m² (large stalls), 3.2 m² (small stalls); Ilkley 4.3 m²; Carzield 2.7 m²; Halton 3.2 m².

⁸⁹ It should not be forgotten that accommodation would also have to be provided for grooms; Speidel, M.P., ‘The soldier's servants’, Ancient Society xx (1989), 239–47.CrossRefGoogle Scholar

⁹⁰ Schröder, op. cit. (note 44).

⁹¹ Crowther below.

⁹² C. Reichmann (in litt.) is of the opinion that the buildings in Period II at Krefeld-Gellep are too large to have housed just the mounts and equipment.

⁹³ cf. Arrian, Ars Tactica.

⁹⁴ Holder, op. cit. (note 38); Dixon and Southern, op. cit. (note 38).

⁹⁵ Josephus, Bell. Jud. III.120; and a number of well-known inscriptiones RIB 254, RIB 481, ILS 2325 (all first-century), and ILS 2326 (second-century); the soldiers themselves may have been accommodated in the barracks either two per century or have remained with the century in which they were first enlisted, Holder, op. cit. (note 38), von Petrikovits, op. cit. (note 37), 53, who notes that they and their mounts need not have been housed together; but the possibility remains that they were, as suggested for Inchtuthil, Pitts and St. Joseph, op. cit. (note 36), 169–71. We can reject any suggestion that the conversion of the buildings in the north-west quarter of Usk was undertaken to provide accommodation for the equites legionis and their mounts, as even if they were not accommodated across the centuries, specialist accommodation for these troops would have lain elsewhere in the fortress. Moreover, any such specialist accommodation would have been provided as a primary consideration during the layout of the fortress, which in the north-west quarter was clearly designed to take cohort blocks. Even if this was not the case, it is clear that the modifications were occurring in two separate cohort-blocks, a space far to ample to accommodate 120 legionary cavalry.

⁹⁶ It is possible the mounts of these officers may have been stabled within their residences, as is certainly evident in some praetoria in auxiliary forts, e.g. Housesteads (Charlesworth, D., ‘The commandant's house, Housesteads’, Arch. Ael.⁵ iii (1975), 17–42)Google Scholar; Loughor (Marvell and Owen-John, op. cit. (note 75)). Provision would also be required for animals used for transport (cf. the discussion by Davison, op. cit. (note 25), 134–6).

⁹⁷ See Holder, op. cit. (note 38); Dixon and Southern, op. cit. (note 38).

⁹⁸ Muller, op. cit. (note 26); C. Reichmann in litt.

⁹⁹ Hod Hill, Richmond, op. cit. (note 88); Longthorpe, Frere and St. Joseph, op. cit. (note 88).

¹⁰⁰ e.g. Bonn, Neuss, Mainz, von Petrikovits, op. cit. (note 37), 56–7 and n. 38. See Dixon and Southern, op. cit. (note 38), 137–44, for a summary of the uses of cavalry with references.

¹⁰¹ Zienkiewicz, op. cit. (note 75), 80–3.

¹⁰² E.B. Birley, ‘Roman garrisons in Wales’, Arch. Camb. cii (1952–3), 9–19, esp. 17–19; V. Nash-Williams, The Roman Frontier in Wales (2nd edn, M.J. Jarret, 1969), 15; CIL XVI.48.

¹⁰³ RIB 201, Camulodunum; Cirencester, J. Wacher and A. McWhirr, Early Roman Occupation at Cirencester (1982), 69 and n. 10.

¹⁰⁴ Zienkiewicz, op. cit. (note 75), 81.

¹⁰⁵ See discussion of site chronology below and Manning, op. cit. (note 4), 45–52.

¹⁰⁶ ibid.

¹⁰⁷ von Petrikovits, op. cit. (note 37), 56–7, these are located in the central scamnum of the praetentura, in the position recommended by Hyginus (von Petrikovits, n. 37). Although the barracks of ala I Thracum have yet to be located at Caerleon, Zienkiewicz places these in the same area as Neuss, op. cit. (note 75), 82–3 and n. 102.

¹⁰⁸ op. cit. (note 75), 83 and nn. 100–1.

¹⁰⁹ The space assigned at Caerleon by Zienkiewicz for ala I Thracum behind the scamnum tribunorum is calculated by him as 1.3 ha (ibid., 83).

¹¹⁰ op. cit. (note 7) 165–66 and fig. 71.

¹¹¹ Seven for the legionary officers – five tribuni angusticlavii, the tribunus laticlavius, and the praefectus castrorum – and an additional house for the praefectus alae. At Caerleon these can be fitted into the scamnum tribunorum, Zienkiewicz, op. cit. (note 75), 83–4 and fig. 26; the possible arrangements at Usk may be closer to those at Neuss.

¹¹² Manning, op. cit. (note 7), 131–5, 173 and figs 4, 56, 59, 73.

¹¹³ ibid., 173–7, figs 4, 7. 73–4> pls XXXI, XXXII.

¹¹⁴ ibid., eastern ditch (one complete section) 3.2 m wide by 2.4 m deep; its line was confirmed for some 52 m through one partially hand-dug and three mechanically-cut sections. Southern ditch (six sections) width 3.2–4.2 m, depth 1.15–1.58 m; the southern ditch has a basal slot in four of the excavated sections with evidence of recutting present in all but one section. Some 45.5 m of the ditch was traced within the Cattle Market site. Given the unusually narrow shape of the fort it is possible that the southern ditch is in fact that of an annexe. This may also explain the earlier closure of this feature and the general absence of any internal features (apart from a possible gate-tower see note 121 below) belonging to the Flavian fort within its bounds on the Cattle Market excavation.

¹¹⁵ ibid., 180.

¹¹⁶ The Flavio-Trajanic roads and ditch here may lie in an annexe.

¹¹⁷ Evans and Metcalf, op. cit. (note 8), 34–5.

¹¹⁸ Manning, op. cit. (note 7), 180–1.

¹¹⁹ Jones, M.J., Roman Fort Defences to AD 117, BAR Brit. Ser. 21 (1975), 69–71Google Scholar and fig. 14 for rampart widths and 105 for berms. Virtually all examples of rampart widths fall in the range 5.5–7.6 m. Jones notes that the most common width for berms is c. 1.5 m and that there are very few examples more than 3 m wide.

¹²⁰ Fig. 16 shows the inner edge at a mid-point of this range.

¹²¹ Manning, op. cit. (note 7), 180, calculated this length as one possibility on the basis of the suggested location of the southern gateway ( ibid., 177–9, fig. 75) and its probable centrepoint and assuming that the eastern and western halves of the fort were laid out on a ratio of 1:1.5. This was, however, based on the premise that the southern side was one of the longer sides of the fort and the position of the gates on these sides is not standardised but set out to known proportions (1:1.5 is evident at Caernarfon and Brecon Gaer). What is clear now, however, is that this is one of the shorter sides of the fort and that either the gate is offset or the post-holes are either part of a tower or not associated with the Flavian fort (they cannot be conclusively associated with it; ibid., 178) and that there may have been no entrance through the southern side.

¹²² Allowing an average ditch width of 3.5 m and berms 2 m wide on all sides.

¹²³ No trace of any features, including, surprisingly, the rampart, apart from the possible gate-tower had survived on the Cattle Market site, Manning, op. cit. (note 7), 173–81.

¹²⁴ Indeed if the intervallum and the via sagularis were the same on all sides of the fort, building space in the interior would be reduced to an elongated strip encompassing an area of c. 0.4–0.5 ha.

¹²⁵ Bishop, op. cit. (note 75), passim.

¹²⁶ op. cit. (note 4), 47; (note 7), 180; see also Mason, D.J.P., ‘Prata Legionis in Britain’, Britannia xix (1998), 163–90, esp. 183.Google Scholar

¹²⁷ Manning, op. cit. (note 7), 180; Mason, op. cit. (note 126), 174.

¹²⁸ See summary below; such evidence was obviously not available to Professor Manning.

¹²⁹ Loughor, Neath, Caerphilly, Penydarren, Coelbren. Abergavenny and Cardiff, both Neronian foundations, continued in use; see P.V. Webster, ‘The Roman Period’, in H.N. Savory (ed.), Glamorgan County History. Volume 11 Early Glamorgan (1984), 277–314; Davies, J.L., ‘Roman Military Deployment in Wales and the Marches from Claudius to the Antonines’, in Hanson, W. F. and Keppie, L.J.F. (eds), Roman Frontier Studies 1979, BAR Inter. ser. 71 (3 vols) (1980), 255–77Google Scholar. For a rejection of Neath as a pre-Flavian foundation see Marvell, A.G. and Heywood, B., ‘Excavations at Neath’, Bull. Board. Celt. Stud. xxxix (1992), 171–298, esp. 174–5, 190, 266, 289.Google Scholar

¹³⁰ Loughor, Marvell and Owen-John, op. cit. (note 75); Neath, Marvell and Heywood, op. cit. (note 129), 289 and n. I; Caerleon, Zienkiewicz, op. cit. (note 75), passim.

¹³¹ Manning, op. cit. (note 7), 180–1.

¹³² ibid.

¹³³ ibid.

¹³⁴ Activity on the nearby 10 Old Market Street site cannot be extended after A.D. 100, and an earlier cessation is likley; Evans and Metcalf, op. cit. (note 8), 34–5.

¹³⁵ Identified by Janet Webster during an initial screening of the material while in conservation at University College, Cardiff, as a Hawkes and Dunning (Med. Arch. v) Type IIA or IIc, cf. G. Clarke et al. The Roman Cemetery at Lankhills, Winchester Studies 3.ii (1979), 286–91, who suggests the date for the Type IIA buckles (of which Type IIc are probably derivative). There is a buckle-plate of a single-strap belt of second-half-of-fourth-century date, J. Webster (forthcoming), No. 113, from Manning's excavations (J. Webster in litt. 19/10/1990).

¹³⁶ See note 2.

¹³⁷ op. cit. (note 4), 24–34.

¹³⁸ ibid., 33; Pitts and St. Joseph, op. cit. (note 36), 276–80.

¹³⁹ e.g. Inchtuthil where the fortress was never completed but the full complement of barracks is present; Pitts and St Joseph, op. cit. (note 36), 276–80.

¹⁴⁰ op. cit. (note 4), 43–52.

¹⁴¹ Manning, op. cit. (note 7), 173.

⁴² See notes 129 and 130 for refs.

⁴³ Zienkiewicz, op. cit. (note 75), 83.

⁴⁴ Marvell and Owen-John, op. cit. (note 75).

⁴⁵ Marvell and Hey wood, op. cit. (note 129), 289 and n. 1.

¹⁴⁶ Davies, op. cit. (note 129).

¹⁴⁷ op. cit. (note 7), 173–81.

¹⁴⁸ See notes 129 and 130 for refs.

¹⁴⁹ See note 135.

¹⁵⁰ op. cit. (note 26), 129.

¹⁵¹ Laboratory work was undertaken by I.B. Clewes and L. Mees.

¹⁵² Dick, W.A. and Tabatabai, M.A., ‘An alkaline oxidation method for the determination of total phosphorus in soils’, Journ Soil Science Soc. America xli (1977), 511–14.CrossRefGoogle Scholar

¹⁵³ Bernas, B., ‘A new method of decomposition and comprehensive analysis of silicates by atomic absorption spectrophotometry’, Analytical Chemistry iv (1968), 1682–6.CrossRefGoogle Scholar

¹⁵⁴ Ball, D.F., ‘Loss-on-ignition as an estimate of organic matter and organic carbon in non-calcareous soils’, Journal of Soil Science xv (1964), 84–92.CrossRefGoogle Scholar

¹⁵⁵ Schröder, in Müller, op. cit. (note 26), 129.

¹⁵⁶ The following abbreviated references are used: RIC – H. Mattingly, E.A. Sydenham, C.H.V. Sutherland and R.A. Carson, Roman Imperial Coinage (10 vols, 1923–1970); BMC – British Museum Catalogue of Coins from the Roman Empire; Cr. – M.H. Crawford, Roman Republican Coinage (2 vols).

¹⁵⁷ M.R. Hull, Roman Colchester (1958), fig. 99, 6.

¹⁵⁸ Hartley, B.R. and Dickinson, B.M., ‘The Samian stamps catalogue’, in Manning, W.H. (ed.), Report on the Excavations at Usk 1965–1976: The Roman Pottery (1993), 210, 47–55.Google Scholar

¹⁵⁹ Durand-Lefebvre, M., Marques de potiers gallo-romains trouvées à Paris (1963), 99, 306.Google Scholar

¹⁶⁰ Hartley and Dickinson, op. cit. (note 158), 210, nos 58–60.

¹⁶¹ F. Hermet, La Graufesenque (Condomatago) (1934), pl. 112, 117c.

¹⁶² Hartley and Dickinson, op. cit (note 158), 213, nos 125–6.

¹⁶³ Simon, H.-G., ‘Die Funde aus dem Bereich der Sumpfbrücke bei Bickenbach (Kreis Darmstadt)’, Saalburg Jahrbuch xxxiv (1977), 52–77, No. 73.Google Scholar

¹⁶⁴ Except where stated all vessels are South Gaulish and come from La Graufesenque. The fabric of the archived vessels has not been checked closely but, as well as vessels from the main South Gaulish production areas and from Les Martres-de-Veyre, vessels from Montans may be present.

¹⁶⁵ See R. Knorr, Töpfer und Fabriken verzierten Terra-Sigillata des ersten Jahrhunderts (1919), Taf. 22.

¹⁶⁶ ibid., 14, 15.

¹⁶⁷ ibid., 3.

¹⁶⁸ R. Knorr, Terra-Sigillata-Gefässe des ersten Jahrhunderts mit Töperfernanem (1952), Taf. 36.

¹⁶⁹ ibid., Taf. 21.

¹⁷⁰ Hartley and Dickinson, op. cit. (note 158), 163, No. 56.

¹⁷¹ Atkinson, D., ‘A hoard of Samian ware from Pompeii’, JRS iv (1914), 27–64.Google Scholar

¹⁷² Hartley and Dickinson, op. cit. (note 158), 162, No. 52.

¹⁷³ Knorr, op. cit. (note 165), 96C.

¹⁷⁴ J.A. Stanfield and G. Simpson, Central Gaulish Potters (1958), fig. 4, 4.

¹⁷⁵ S. Rogers, Poteries sigillées de la Gaule Centrale, Gallia Suppl. 28 (1974), 63.

¹⁷⁶ ibid., probably R91.

¹⁷⁷ Hartley and Dickinson, op. cit. (note 158), n. 151.

¹⁷⁸ For a discussion of the lagena see Balsan, L. and Vernhet, A., Gallia xxix (1971).Google Scholar

¹⁷⁹ Hermet, op. cit. (note 161), pl. 28, 68.

¹⁸⁰ M.C. Bishop and J. Dore, Corbridge. Excavation of the Roman Fort and Town 1947–1980, HBMCE Arch. Rep. 8 (1988), No. 8.

¹⁸¹ Knorr, op. cit. (note 165), Taf. 36C.

¹⁸² ibid., Taf. 36A.

¹⁸³ Atkinson, op. cit. (note 171).

¹⁸⁴ Knorr, op. cit. (note 168), Taf. 53.

¹⁸⁵ M. Henig, A Corpus of Roman Engraved Gemstones from British Sites, BAR Brit. Ser. 8 (2nd edn, 1978), 35, fig. 1, shape A3.

¹⁸⁶ For the type see M. Henig and M. Whiting, Engraved Gems from Gadara in Jordan. The Sa'd Collection of Intaglios and Cameos, Oxford Univ. Comm. Arch. monograph 6 (1987), No. 234 (cornelian from Gadara, type as Usk intaglio; also No. 235 with column in field); Gozenbach, V. von, ‘Römische Gemmen aus Vindonissa’, Zeitschrift für Schweiz. Arch. und Kunstgeschichte xiii, 67 No. 6Google Scholar, pls 27 and 29 (cornelian from Vindonissa, type as Usk gem but to left); G. Sena Chiesa, Gemme del Muzeo Nazionale di Aquileia (1966), No. 500 (cornelian from Aquileia; column, with ?urn in field); Maddioli, G., ‘Le cretule del Nomophylakion di Cirene’, Annuario delta Scuola Archeologica di Atene xli–xlii, 113 No. 678Google Scholar, fig. 34 (sealing from the Record Office Cyrene, destroyed in Trajan's reign; column with urn in field). The two Dioscuri, each before the foreparts of his mount, are seen, for example, on a cornelian in The Hague: M. Maaskant-Kleibrink, Catalogue of the Engraved Gems in the Royal Coin Cabinet The Hague. The Greek, Etruscan and Roman Collections (1978), No. 824. Other intaglios figure a Dioscurus with his horse completely visible, as Sena Chiesa, loc. cit., Nos 496–9; E. Brandt, Antike Gemmen in Deutschen Sammlungen I. Staatliche Münzsammlung München, 3. Gemmen und Glaspasten den römischen Kaiserzeit (1972), No. 2727.

¹⁸⁷ Henig, op. cit. (note 185), no. 510.

¹⁸⁸ ibid., No. 97. For standing figures of the Dioscuri without their horses, all on intaglios of the later second or even the third century, ibid., Nos 95 (Silchester) and 96 (Pentre, Rhondda, Glamorgan); J.D. Zienkiewicz, ‘The Engraved Gemstones’, in idem, The Legionary Fortress Baths at Caerleon II. The Finds (1986), 136, No. 49 (Caerleon).

¹⁸⁹ See Gury, F., ‘Dioskouroi/Castores’, in Lexicon Iconographicum Mythologiae Classicae III (1986), 608–35.Google Scholar

¹⁹⁰ Henig, M., Webster, G. and Wilkins, R., ‘A bronze Dioscurus from Wroxester and its fellow from Canterbury’, Antiq. Journ. lxvii (1987), 360–2Google Scholar. There is another example in Colchester.

¹⁹¹ Archaeologia lviii (1903), 573Google Scholar, pl. 55.

¹⁹² J.M.C. Toynbee, Art in Roman Britain (1962), 168, No. 104, pl. 110.