Though the limits of 17220 were clearer than 17216, we still frequently defined and clarified these boundaries through the testing of hammerscale levels with a magnet. The cut containing 17220 was excavated until both the darker grey deposit was removed and none of the cut walls were attracting the magnet. Stratigraphic reliability is moderate here because we did not identify the surface into which 17220 appeared to be cut. The square cut containing 17220 emerged below the level to which 17215 was excavated, where we had artificially taken 17215 down in order to find the limits of the forge deposit (thus identifying cut 17225 and deposit 17220).

Deposit composition is forge scale mixed with silty clay, but primarily (80%) the deposit comprises forge scale. We went back and forth over the forge feature (comprising deposits 17216 and 17220, as well as cut 17225) from the beginning to end of its excavation (12-20 July). Below is a summary of the observations presented chronologically with dates in parentheses so that photos can be matched up with descriptions. Originally 17216 was differentiated from the covering and surrounding deposits (17211 and 17215 respectively) by several pockets of dense magnetism in the centre of a tile cluster (generated by hammerscale and iron fragments). Magnetism was moderate surrounding this immediate cluster extending about 30cm from the boundary of the tiles (14/7). Upon excavation and removal of the tile and brick concentration an oblong oval of hammerscale surrounded by a gray clay (17215) was identified oriented E-W. It contained slag, bronze fragments, and iron fragments. We took this to be the beginning of the forge deposit and rerecorded the boundaries of 17216 as the extent of this oval (morning 15/7). As the hammerscale was removed from the oval, it became clear that the hammerscale lens began to slope under 17215, especially towards the northern baulk (end of day 15/7). At this point we theorised that the ovular concentration might be an accumulation of metalworking refuse on top of a surface or a filled cut below, because 17216 undercuts 17215 in the north in a way that undermines the original interpretation of 17216 as cut-fill. As a result, we decided to level the area by stripping 17215 back to create an artificial square sondage (18/7) and we found a large square cut containing a hammerscale deposit that could be differentiated through colour (darker grey), compaction (mod compact) and composition (dense concentration of hammerscale gave the matrix a sandy texture, though the soil itself is silty clay). 17216 quickly expanded in its limits to cover a square-ish area of roughly 60cmx70cm, and continued into the baulk wall which was stripped back until it no longer attracted the magnet. As can be seen in the basic sketch plan of the linked SU sheet, the square cut was multifaceted in composition and featured some layering. The footprint of the gray oval, which had been seen at the beginning of 17216, continued to present dense concentrations of hammerscale and forge refuse in the square cut. After excavating some of this deposit, a clay mound emerged in the south-eastern portion of the square cut (18/7). Hammerscale continued to come up between this clay mound and the baulk. After the hammerscale was further excavated it began to dive down below the level of the clay mound towards the baulk. We considered assigning the two areas (hammerscale and clay mound) different SUs, but decided against it once we dug into the clay a little and it became clear that the clay mound was more of a lens. At this point we changed SU across the whole deposit from 17216 to 17220. Upon the excavation of the clay lens there was an initial layer of hammerscale, and then a layer of slag appeared (19/7). This slag was the type found at the bottom of forges and welding droplets were still present. After the slag layer was removed, a layer of a light gray ash was observed, and below that was a layer of gray clay forming a shallow basin. This feature had all aspects of being a small forge, although it was different in construction from forges found in prior years. It was also at this point that a high density of nail heads was found, both in and around the gray clay depression, and in the continued deposit of hammerscale north of the clay depression and south of the baulk. Due to the gray clay basin sitting on top of the continuation of the hammerscale deposit, it was removed to further explore the extent of the metal working refuse. Nail heads and iron fragments remained in high frequency throughout the remainder of the hammerscale deposit, while the frequency of slag decreased. The cut eventually terminated on a dark redish gray clay which lined the bottom (20/7). Once 17220 was fully removed, the cut was assigned the SU number 17225. With retrospect it is reasonable that 17216 and 17220 could be understood as separate sections of the forge deposit: 17220 could be classified as the deposit that filled a square depression that was cut into the floor for the purposes of metalworking, whereas 17216 is a deposit that accumulated on top of this filled cut. However, since 17220 was assigned after excavation of the square cut had already begun and special finds are categorised accordingly, I think it is better to take 17216 and 17220 together as ‘forge deposit’ and understand that there is layering and phasing between and within these units. Ian interprets the clay depression in 17220 that contained a high amount of slag as the bottom of a forge. That hammerscale continues to come up underneath this feature suggests metalworking that preceded this forge.