Up to the mid-fourth century AD, the language of refuge regularly appears in Roman sources in the context of frontier management. It is employed both of high status individuals, but also – more strikingly – of very much larger groups: certainly several tens of thousands of individuals, and sometimes apparently a hundred thousand plus-strong. The basic political economy of the Empire – powered by unmechanised agricultural production in a world of low overall population densities – meant that there was always a demand for labour, and, in the right circumstances, refugees could expect reasonable treatment. Provided that their arrival posed no military or political threat to imperial integrity, refugees would receive not only lands to cultivate on reasonable terms, but might also be settled in concentrations large enough to preserve structures of broader familial and even cultural identity. In other circumstances, however, imperial control was enforced by direct military action and survivors were sold into slavery and might themselves redistributed as individuals in adverse socio-economic conditions over very wide geographical areas. In the late fourth and early fifth centuries, a distinct change becomes apparent in imperial policy. Some very large refugee groups – particularly those that were Gothic – were granted lands within the Empire on terms which broke with long-established Roman norms. These groups were so large and retained so much autonomy that they posed a distinct threat to the continued integrity of imperial rule over the particular regions in which they were settled. Over time, some of the settlements eventually became the basis of independent successor kingdoms as the power of the west Roman centre unravelled. This transition poses an obvious question. Why did traditional Roman policy towards refugees change so markedly in the late imperial period?