Excavation was undertaken beneath the floor of Castle House in 2015 while construction work was still in progress, with three rectangular holes for glazed viewing panels built into the modern concrete floor. The aim of the 2015 excavation was to expose and record the remains of the Roman arcade beneath the modern concrete floor, so that the remains would be visible through three viewing panels (the ground floor was raised significantly above the archaeological remains). The natural subsoil was not reached during the investigations in 2015; it is c.2.6m from the top of the modern concerete floor based on earlier investigations. The foundation platform (F2) that supported the Roman arcade was uncovered for approximately 12.7m E to W; only the upper surface of the northern part of the foundation platform and its northern edge were uncovered in 2015. It was constructed of a hard, pale brown mortar containing fragments of septaria. There was evidence that the northern part of the foundation platform (F2) was resurfaced during the Roman period, with a layer of opus signinum set on a bedding layer of rubble in pale brown mortar. The remains of three piers (F17, F3 and F12) and four blocking or revetting walls (F4, F13, F14 & F20) were exposed on top of the foundation platform. The remains of the three piers extended for c.0.60-0.80m above the top of the foundation platform F2 (1.0-1.2m below the top of the modern concrete slab). Only the cores of the piers survived, as the facing stones had been almost completely robbed out, probably in Norman times. The best preserved of the three piers excavated in 2015 was the middle one (F3). This was fully uncovered in 1964 and was re-exposed in 2015. The remains of this pier consisted of a hard stone-and-mortar core that survived approximately 0.80m above the level of the top of the foundation platform F2. It measured c.1.06m E to W x 0.90m N to S. On the east side of F3, a small piece of facing stone, identified as Ham Hill stone, survived in situ. The revetting walls (probably intended to counteract the build-up of deposits on the south side of the arcade) survived between 0.60m and 0.90m high (approximately 1.0-1.3m below the top of the modern concrete slab). F4, F13 & F14 lay within the limits of the 1964 excavation. The revetting walls F4, F13 and F20 were approximately 0.80m wide, while F14 was slightly narrower with a width of approximately 0.74m. It was possible to record the full lengths of revetting walls F13 and F14, which were 2.25m and 2.0m long respectively. Their northern faces consisted of courses of septaria and Roman brick/tile. The southern faces of the revetting walls were coated with layers of opus signinum, plaster and mortar rendering. Where visible, the cores of the revetting walls consisted of pale brown mortar containing fragments of septaria and pieces of Roman brick/tile, with some gravels. A sequence of brownish, gravelly layers (L26, L24, L23 & L18), of probable Roman date, was excavated in the south-eastern corner of the site, surviving in a narrow strip to the south of the easternmost Roman revetting wall (F20). Thick deposits of greyish-brown dereliction and/or demolition debris that derived mainly from the Roman arcade were found on the site. They contained large quantities of loose mortar fragments, as well as other pieces of Roman building materials, such as brick, tile, stone, opus signinum, and plaster. Most of the stone was septaria. Also in these deposits were some larger fragments of masonry that were probably the remains of collapsed piers with attached columns. Part of a collapsed attached column (F18) was uncovered, overlying the foundation platform F2. The remains of the collapsed column were approximately 0.90m long, 0.50m wide and 0.30-0.40m high. The column fragment consisted of courses of Roman column brick and septaria set in hard, pale brown mortar. The collapsed column fragment F18 presumably fell from the pier that projected from pier stump F17, which lies a metre or so to the south. Another collapsed fragment (F8) that probably came from the same column as F18 was uncovered further north in 2014 (ECC4219). Approximately 3.50m to the east of F18, a large fragment of Roman masonry (F23) overlay the northern edge of the foundation platform F2 and was associated with thick deposits of demolition debris (L29 & L27). This fragment was probably part of a collapsed Roman pier and/or attached column, lying face down, relating to pier stump F3. A post-Roman feature (F22), probably a gully that extended E to W, contained a small quantity of late 11th- to 12th-century potsherds (but was possibly earlier in date). It lay between the pier F17 and the collapsed column F18. Approximately 3.50m to the east of F18, a large fragment of Roman masonry (F23), probably part of a collapsed Roman pier and/or attached column (that projected from pier stump F3), overlay the northern edge of the foundation platform F2. It consisted of fragments of septaria and Roman brick/tile set in pale brown mortar; traces of a possible face, lying face down, were discernible in the north-western corner of F23. The fragment of collapsed column/pier (F23) was associated with thick deposits of demolition debris (L29 & L27). A quantity of 11th- to 12th-century potsherds was recovered from L27, as well as some residual Roman sherds. Among a group of Roman brick/tile fragments (89) in L27 were two joining Roman brick fragments with an unusual chamfered edge, perhaps derived from an entablature. A sherd of Thetford-type ware and a sherd of late 11th- to early 12th-century date were also recovered from L16, demolition debris, up to 0.40m thick, to the west of the modern north-south foundation F11/F15. Much of the 2015 site had been excavated previously in 1964, so most of the archaeological deposits (and demoltion debris) that overlay the arcade had already been removed. However, undisturbed deposits survived in a few places (along the northern and eastern edges of the site, where these lay beyond the limits of the 1964 excavation), and these were recorded and, where necessary, excavated. Thick deposits of demolition and/or robbing debris survived in situ in the northern part of the site and these were investigated in T3 and T4 in 2014 (see ECC4219). However, a couple of features (F16 & F19) and a layer (L13) were excavated at the eastern end of the site in 2015. They lay to the east of a modern north-south foundation (F11/F15) and beyond the area excavated in 1964. These were probably medieval in date, but postdated the layers of Norman demolition debris associated with the arcade. The demolition debris L17 was cut by an oval-shaped pit (F19) and a small quantity of late 11th- to 12th-century potsherds was recovered from the pit. L13, which sealed F19, also contained a late 11th- to 12th-century sherd. In the south-east corner of the site, L13 was cut by a shallow pit (F16). A small quantity of medieval potsherds was recovered from the fill, and these suggest a late 12th- to early 13th-century or later date for F16. A small copper-alloy mount (SF7) also came from F16. Little definite evidence was found for the Norman inner bailey rampart in either 2015 or in 1964. If, as expected, it did extend over the remains of the arcade, it must have been truncated as a result of later activity.<1><2> This work followed an evaluation (T1 & T2) in 2012 (ECC2988).<3>