as a large oval pit measuring 3.85m (E-W) by 3.50m (N-S). At a depth of 2.20m, however, the main pit narrowed into a vertical shaft, sub-rectangular in plan and measuring 0.80m by 1.05m. A specialist team was able to excavate this shaft for 2.40m when the top of the water-table was reached and safety considerations precluded any further work. No traces of any original lining survived within the shaft but the occurrence of vertical grooves at each corner would be consistent with the former presence of a timber plank lining of typical Roman construction.
   The well had been deliberately backfilled with large amounts of pottery and other domestic rubbish, together with quantities of tile and building rubble. Over 1,200 pot-sherds were recovered. The bulk of this material dates to the second century, with some residual pieces from the first century and a few sherds which could be as late as the early third century (see below, Assemblage D). A Celtic coin of Eppillus (c.AD 1-15) was also recovered (Holman and Parfitt 2005, coin ref. 2003/25; see below), which must be contemporary with the earliest dated (residual) pottery.
   About 1.50m to the north-east of F. 7105, the top of a slightly larger shaft was partially investigated (Fig. 2, F. 7128). It seems fairly certain that this represents a second well, although it could not be fully excavated. The shaft was half-sectioned to a depth of about 2.00m and like F. 7105 was also found to have been backfilled with large amounts of domestic rubbish, tile and rubble. More than 2,800 pot-sherds were recovered from its filling. As with the other shaft, most of the pottery dates to the second century with some residual pieces of the first century (see below, Assemblage E). A number of metal items also came from the filling of this feature. These included objects of copper-alloy and lead, together with iron nail fragments. Of particular importance were three similar, leaf-shaped, copper-alloy pendants. These have been identified as pieces of military equipment (Chris Tucker pers. com.) and are presently being studied in detail. Eventual subsidence of the shaft filling caused a surface hollow to form. This had been deliberately infilled with laid cobbles and tile fragments.
   Given the close proximity of a natural stream in the little valley below, the discovery of at least one definite well on the hill-top occupied by the villa complex is of particular interest. There seems little doubt that the close proximity of the bath-suite in Building 6A, with its requirement for large amounts of water, had dictated the positioning of the well, although nothing similar was identified in the area of the Building 3 baths (Parfitt in Perkins 2004). This may perhaps be explained by the fact that the area of the northern baths stood at an elevation about 3.50m above that of Building 6A, so that consequently any well there would need to be deeper and be more difficult to excavate.
   If, as appears likely, F. 7128 was a second well, it seems most probable that these were successive shafts supplying water for Building 6A.