This is not just an excavation report of a Roman fort in Bulgaria, rather it is an especially intriguing and revealing late Roman fort with surprising finds, including cowbells, scythes, spindle whorls, loom weights, quantities of burnt grain and a ballista bolt bent in action. Behind lofty, well-constructed walls (over 9m) built in the Constantinopolitan manner, with mortared stone and brick in bands, were ranks of mud-brick, two-storey buildings on stone socles (plinths) with a central street leading from a narrow gate to granaries. The site of Dichin is set in the Danube hinterland of the late Roman province of Moesia II. The excavator Andrew Poulter has dedicated his career to the investigation of urban and rural communities of Roman Moesia. Moving on from the successful excavation of late Roman Nicopolis ad Istrum, the main local town, Poulter set out with his Bulgarian colleague Ventzislav Dinchev to find a contemporaneous village site as a comparator. In the foothills of the Balkan mountains many fortified settlements date from this period, but their precise role remains enigmatic: were they defended villages, garrisoned settlements or refuges? The excavations revealed site finds of 406 coins, ranging in date from the end of the fourth century to the late sixth. Coin supply and usage indicate a military presence, although no ancient name is recorded. The unusual fortification, 1.8 hectares in area, is situated within a loop of the River Rositsa, south of Nicopolis and close to the line of the city's aqueduct. It does not lie on a known communication route, but the river provided access to the city and the Danube frontier garrisons to the north.

The destruction of the eastern field army in AD 378 by the migrant Goths at Adrianople (Edirne) together with the death of the Emperor Valens signalled the failure of the Danube frontier to secure the interior provinces and the capital, Constantinople. The new fort at Dichin post-dates that disaster and, through a careful assessment of the coin distribution, Peter Guest suggests a date after 400. Poulter links the construction with an imperial rescript concerning restoration in Moesia II in 408. The fortifications are unusual for such a small site. Dobri Dyal, a fortified hilltop 15km east and excavated by the same team (Poulter Reference Poulter2013), also has mortared walls but without the careful construction seen at Dichin, which includes regular bands of brickwork, buttresses and internal stairs. Such construction is comparable with other major Late Antique cities such as Hisar (Diocletianopolis) or Plovdiv (Philipoppolis), showing that Dichin was defended in the manner of a Thracian city or a frontier fort. Inside the walls were regular rows of buildings varying in size but mostly around 5m wide and between 10 and 16m in length. Steps led up to second floors where the living accommodation was situated, with byres and stores below. Towards the west end, a range of buildings were shown to be granaries with raised floors supported on mud-brick piers. What is unusual in the main occupation period (period 2) is the absence of any recognisable command and administrative centre, common in Roman forts and still found in the Danube forts at this period. Across the west side, an additional outer wall (proteichisma) was added later in the fifth century, very close to the main curtain, as a further defence. The main wall incorporated some spolia, including blocks shown to be from the nearby aqueduct, which had ceased to function by this time. Period 2 came to an end around 500 when there was a major fire across the interior, after which the site was radically rebuilt over the demolished remains of the mud-brick buildings but to a similar plan. The demolition levels produced the richest haul of stratified finds and ceramics, which provide remarkable insights into life in the fifth-century fort.

At Dichin the pattern of much of the interior was revealed through a network of robber trenches, and the results of six seasons of excavation present not just detailed descriptions of the structures but also associated finds. The main specialist reports will provide a crucial datum for future comparisons and research in the late Roman world. The coins are rigorously and assiduously analysed to present a model report. The range of metalwork reads like a fifth-century ironmonger's manifest, from weapons (not so many but indicative of conflict) to scythes, ploughs and a range of other agricultural implements, including cowbells. In this context, cowbells surely represent grazing dairy cattle not just draught oxen as suggested, and perhaps are indicative of cheese production. The finds also hint at the character of the garrison. Spinning and weaving suggests women and families, finds rarely found on third- and fourth-century forts on Hadrian's Wall. Poulter argues that the fort was occupied by foederati—federates drawn from the Gothic settlers—in part based on the absence both of a church in period 2, and a recognisable headquarters structure. Some parts of the interior of the fort remain uninvestigated, however, and there are Christian finds in this phase such as glass and tiles. Neglected in most discussions of the Roman army in the region are attempts to identify the presence of the regional field army that is known to have been restored by the early fifth century. The field army's location remains enigmatic, but it is possible that a combined agricultural production centre and military garrison may have addressed its needs through the system of the Annona, a taxation in kind, rather than serving frontier troops on the river. The fire destroyed significant quantities of grain, but some was preserved. There seems little doubt that we are dealing with soldier-farmers in this context, but not impoverished peasants: there was still a demand for imported ceramics, they ate off metal dishes and the local economy appears to have been monetised. A major discovery from the destruction deposits in periods 2 and 3 (c. 580) was that analysis of burnt grain demonstrates continuing cultivation for a wide range of plant species into the late sixth century, questioning the historical view of collapse and dearth. The author and his team of collaborators are to be congratulated for contributing not just a report and challenging analysis from their fieldwork, but for creating a major and enduring resource for future comparative research into the material culture and history of the region.