The placement and dimensions of the fourth trench (24 x 15 m) were such as to include the northeastern corner of the peristyle and part of a street to its north, both identified thanks to the geophysical prospection. The area slopes gently from west to east. For practical purposes we divided it in three sections of equal size, each measuring 8 x 15 m. At the southern half of the trench the removal of the surface layers (Context 3000), which are successive natural fills from the erosion of the slope, exposed Walls 3001-3002 and 3003-3006 which join at right angle and belong to the complex identified from geophysics. Walls 3001-3002, 1.2 m thick, are built in opus caementicium (rubble, fragments of tile and mortar) lined with semi-dressed small stones and mortar. Walls 3003-3006, are of similar construction, but are thinner (0.9 – 1 m) and their facing seems to be less careful. At the southwestern corner of the trench, the removal of the top layer uncovered a whitish, relatively hard surface (Context 3005) which we decided to keep at least until we extend the trench southwards and westwards during the next season. Excavation between Walls 3001-3002 and 3003-3006, that is in the interior of the supposed portico, brought to light three parallel clay pipes (Context 3012) near the interior side of Wall 3006. The preserved sections of the pipes, which extend from west-northwest towards the east, have an overall length of ca. 4 m and are made of terracotta tubes, 0.57 m long and 12 to 14 cm in diameter. Along the southern side of these channels, three aligned stones are preserved, which may belong to a protective or retaining wall of the channels. The destination of the channels is unknown but towards the west-northwest at least one goes under Wall 3006 and continues under the semi-circular exedra as shown by the short section found at the inner corner of the exedra (Context 3028).
We preserved the layer of the channels and excavated lower in the rest of the area defined by Walls 3003-3006 and 3001-3002. From the thick deposit excavated in 2014 we recovered limited amount of pottery, dating primarily from the Late Roman period mixed with a few Hellenistic and Early Roman sherds. In lower levels, the excavation of Contexts 3026 and 3027 brought to light the upper, level surface of an ashlar block, 1 x 0.75 m, set between Walls 3006 and 3002. A smaller stone is attached to its southern side. These two stones have not been exposed through their entire thickness, and consequently their purpose is still uncertain. It may be a base of a column (part of an inner colonnade) or what is left from an original, paved floor, but only the continuation of the excavation can clarify the situation. Further to the east, we discovered a small section of terracotta channel (Context 3032), 2.3 m long, which is cut by a small layer of rubble and tiles. The channel is oriented northwest-southeast and is made of terracotta pipes of different length but constant diameter of 12.5 cm. The diameter of this channel is similar to the one of the channels that we located near Wall 3006 but the two are surely unrelated since their elevation is 32 cm apart.
Whether the “outer” corner (Walls 3003-3006) and the “inner” corner (Walls 3001-3002) found in the southern half of the trench are part of the same building has not been determined yet, but it is highly likely based on the masonry and their topographical relationship. When these walls were built is still uncertain, and only the excavation of deeper strata can provide an answer. The lower part of the revetment of the northern side of Wall 3002, and the level of the pipes next to Wall 3006 are indicative of the level of a floor made of packed earth which, however, was not found. One possibility is that this floor was level with the hard surface located at the southwestern corner of the trench, that is only a few centimeters below modern ground surface, and therefore destroyed by plowing activity. We certainly had at least one more floor level at lower levels, perhaps related to the stone that we found in situ between Walls 3006 and 3002, even though we found no destruction layer associated with it. To a later phase most likely belongs the semi-circular exedra (Context 3007) since its extremities are not bonded with Wall 3006 and the uppermost preserved surface of Wall 3006 (ca. 167.5 masl) bears no traces of a threshold which means that the entrance to the exedra was at a still higher level. This is a further indication that the surface of packed earth at the southwestern corner of the trench (167.67 masl) may have also been the floor surface in the last phase of the building.
The semi-circular exedra which opens onto Wall 3006 (Context 3007), with a maximum radius of 2.7 m, is built with rubble, tile pieces, and mortar, faced in opus incertum. The average thickness (0.90 m) and the masonry are similar to those of Wall 3006 but their contact shows that the exedra is a later addition. The layers excavated inside the exedra (Contexts 3008 and 3014) contained substantial amounts of pottery mostly of the Late Roman period but mixed with some earlier sherds. After excavating these layers we discovered the terracotta channel (Context 3028) at the southwestern corner of the exedra and part of an earlier wall (Context 3022), 0.45 m thick, parallel to Wall 3006. In the masonry of the wall, which is built carelessly with stones of different dimensions and without mortar, a few architectural members are visible including a marble double column set horizontally. It is possible that the visible section of this wall is a later addition on top of earlier foundations preserved lower. At any rate, this wall was built earlier than the exedra. The material from the collapse of the upper structure of the exedra was found in a thick layer immediately east of the exedra (Context 3009) along with ceramic material dating down to the 7th century CE. Inside this layer we also found part of a column and an ashlar block which, together with two more column fragments found in the northern trench, likely come from Wall 3017.
The surface layer to the north of the exedra (Context 3010) surpassed 1 m in thickness and contained material of many periods, from the Early Hellenistic to the modern. After we took out this layer, we brought to light the outline of three walls (Contexts 3015-3017). The northernmost wall (Context 3015), oriented east-west, is preserved over a length of 7.8 m and consists of a course of ashlar blocks set on their short side, manifestly in secondary use. The preserved section is not entirely straight: some stones are dislocated, perhaps on account of the weight of the earth that they withheld, and others are tilted northwards. Directly north of this wall and some 40 cm below the ground surface, we discovered a hard surface with traces of whitish mortar which may belong to a later phase of the street that we are seeking but without continuation (and extension) of the excavation this cannot be ascertained. If this hypothesis holds, then Wall 3015 would function as the retaining wall of the street.
Parallel to Wall 3015, towards the south, is Wall 3016, 6.83 m long, of which the western end joins Wall 3017 at a right angle. These two walls (Contexts 3016 and 3017) are obviously contemporary and built mostly with spolia among which we can single out part of a semi-fluted column, a triplyph-metope piece of large dimensions (1.41 m long, 0.46 m wide, and 0.92 m high), a second, smaller triglyph-metope piece, a fragment of a triglyph, part of geison and part of an epistyle. Given that the southern end of the Wall 3017 touches the apse of the complex (Context 3007) it is conceivable that this wall, 3.15 m long, was built in order to prop the northern face of the apse and/or the southern face of Wall 3016. It is worth mentioning that only the western half of the external face of the exedra is plastered with zones of incised double lines forming a cross-hatching motif, and that the dividing line is exactly where Wall 3017 joins the exedra. In other words the plaster was put into place after the construction of Wall 3017 (and 3016).
The layers that we excavated (Contexts 3011, 3020, 3023) between the line of Wall 3015 to the north, and the apse (Context 3007) and Walls 3017 and 3016 to the south, 0.80 m thick, yielded abundant ceramic material mostly but not purely of Late Roman times. From the fill layers (Contexts 3024 and 3030) between Wall 3016 to the north, Wall 3017 to the west, and the corner of the complex to the south, come the spolia mentioned above (and registered as 3024.01, 02, 03, 04). The dating of the construction of Wall 3016-3017 cannot be determined yet and its use remains puzzling. It may have acted along with its parallel Wall 3015 (the distance between the two is ca. 1.7 cm), as a system of encasement of a small stream coming from the upper plateau but only the extension of the trench to the west and the east can check such a hypothesis. What is certain is that the construction of these walls (3015-3017) together with the plastering of part of the external face of the apse (Context 3007) represent the last building phase in the northern side of the trench. Their builders made extensive use of earlier material coming from public buildings, perhaps from one of the stoas that existed here (to which the half-fluted column certainly belonged) as well as from the excavated Hellenistic temple from which the large triglyph-metope piece is likely to come.