There are good indications that the earthwork at Castle Flemish enclosed a Roman villa, or romanised farmstead, probably including a bathhouse, occupied from the late first century AD onwards. Legends of a golden table buried on the site have not been disproved.

It is an irregular quadrilateral banked enclosure, about 90m east-west by 82m, round-angled and straight sided, set on ground falling gently to the south. The site was identified with the spurious 'Ad Vigessimum' of Richard of Cirencester; brick was noted here through the nineteenth century, along with persistent reports of inscribed stones and of a golden table. Trenching and pitting over three days in December of 1922, determined that the greater part of the interior had been cleared and levelled, most test-pits producing only fragments of brick and slate. A section taken across the bank close to the south-west angle showed that this was some 6.0m wide, survived to 1.0m high, and was separated from a ditch, some 4.0m wide and at least 2.1m deep, by a 1.0m wide berm. The front of the rampart is thought to have been stone kerbed, or revetted. A trench in the south-eastern part of the enclosure recorded a sequence of two clay floors and their substructures, about 0.48m of Roman stratigraphy being recorded; the upper floor may have extended over an area of up to 9.0m overall, with evidence of a hearth, or partition, and was associated with hexagonal roofing slates; a later first to earlier second century ceramic assemblage predated this later floor. Unstratified finds of flue-tile fragments imply the presence of a heated appartment, presumably part of a bathhouse.

A bank crossing the site (NPRN 23710) appears to represent a recent field bank, rather than a Roman road.

Source: Wheeler 1923 (AC 78), 211-224.

J.Wiles, RCAHMW, 14 February 2005