Because of the outbreak of the Second World War the extensive archaeological work undertaken by Drew and Collingwood Selby at Colliton Park in 1937–8 was never brought to publication. Nearly 70 years later Emma Durham, Mike Fulford and their colleagues have rescued what can be rescued relating both to the history of occupation on the site and the large haul of artefactual material. This report was assembled over 30 years on a series of limited budgets. That the two principal authors and their collaborators have managed to extract as much meaning as they have from what is left is a tribute to their persistence.

The excavations at Colliton Park started in the year of the last season of Mortimer Wheeler's campaign at nearby Maiden Castle. The excavation methodology was clearly inspired by his example and his support, consisting ultimately of a grid of 25-feet-square ‘boxes’ being laid out over the 10-acre (4-ha) site. Some ‘boxes’ remained unexcavated, some were only trenched: where structural remains were encountered an area was opened up to clear the full plan. The numbering system of Buildings devised by the RCHM and translated to Monuments by the HER is not an aid to comprehension. Deficiencies in recording and subsequent loss of information mean that most finds are now effectively unstratified. The report therefore concentrates on the structural sequence and on presenting the principal classes of finds, save the pottery.

Occupation on the site started in the later first century with timber structures, ditches and pits and a major culvert running downslope. Through the second and into the third century there seems to have been a low level of activity, possibly including the dumping of material from elsewhere. It was only in the later third and fourth centuries that this peripheral area of the city saw sustained activity with the construction of several structures on flint and stone foundations, two being particularly notable. One, Building III in Monument 84, consisted of a long (48 m), narrow, north–south range of apparently artisan or agricultural function. The other, Building 182, was recognised on excavation as remarkable and provided with cover-buildings, renewed in 2007. Three separate three-roomed buildings probably constructed in the early fourth century were amalgamated into a single building by linking corridors in the middle of the century, the differing alignments of the original three units giving a very irregular plan. One unit, the South range, seems to have been a ‘service range’, the other two, now the West range, were floored throughout in polychrome mosaic. The higgledy-piggledy plan is hard to parallel but does recall the larger and more elaborate Insula XIV.2 house at Silchester, also an amalgamation of pre-existing units. Stephen Cosh gives an extended and thoughtful consideration of the mosaics and their comparanda. There was also a quantity of polychrome plaster (reported on by Jane Timby), including the use of the expensive pigment cinnabar. Because the plan deviates so much from the more usual axial layout of late Roman residences and because none of the rooms has a distinctive plan, it is not possible to reconstruct room uses. Nevertheless, with these mosaics and wall-plaster, the rooms must have been intensely colourful, lit either by restricted daylight (note the remains of a window embrasure from Room 10) or by the candles held in the tripod candlesticks found in the building. The rooms range from 3.3 m by 2.8 m to 6 m by 5.4 m, and once one imagines them populated with furniture and soft furnishings (in various colours) they must have imparted a very specific physical and visual experience.

There was a large quantity of finds from the site, unfortunately now mostly lacking location and context. Because of this and the limited means of the project, of the ceramics only the samian is reported on here. Given the history of the site the emphasis in datable finds is towards the later Roman period, though there is a quantity of clearly earlier material in both the samian and the metal finds. An exceptional find is a fourth-century glass bowl with incised Bacchic decoration, which someone had unsuccessfully tried to cut in two (why?). An important resource for the study of a local industry is the quantity of pieces of Kimmeridge shale, most notably the complete table-leg, but also a fragment of a vessel with a graffito in Greek (sadly incomprehensible). The group includes manufacturing debris, adding to the evidence that shale was worked at Dorchester. There are also two coins unique in Britain, silver miliarenses of Constans, buried and not recovered beneath a floor in a poorly preserved house, Building 186. Was this something to do with Britain's part in the overthrow of Constans? In general the coins show low numbers until the later third century with a higher presence thereafter, a pattern often found on rural sites. On the other hand the sheer volume of artefactual material from the site is very much an urban signature, so the coin profile probably owes more to the occupation history of this peripheral location than to the site's function. Of interest is the presence of some military-related objects from the second century onward, part of the growing evidence that soldiers may have been a familiar sight in the towns of Roman Britain, even this far south.

That one can pick up so much of interest from the report despite the problems of the excavation, recording and curating of the site and its material is a tribute to the work of the collaborators in this project, who deserve our grateful recognition.