The reforms of the Roman army by Diocletian and Constantine separated it from the civil administration, so that governors of provinces no longer commanded a provincial army, although they were still responsible for raising funds to support it. The army was divided into two parts: there were troops protecting the borders, the limitanei, under the command of duces; and there was a field army, the comitatenses, which was mobile and organised in divisions under the command of the magistri militum. In addition there were the palace troops and the imperial bodyguard, whose titles changed throughout the fifth and sixth centuries. By the ninth century a quite different system had emerged, with the army divided into divisions called themes, based in provinces also called themes (themata), each under the command of a strat¯egos. There is no general agreement about how quickly this change took place, or why: was it the result of some planned reorganisation, or simply a fumbling reaction to the problems of the seventh and eighth centuries? There is, however, general dissent from the theory which once commanded much support, and is associated with the name of the greatByzantinistGeorgeOstrogorsky, which saw the thematic army as the result of a deliberate reorganisation of army and empire by Heraclius. The result, supposedly, was a peasant army, based in the themes, in which land had been allotted to peasant families as small holdings, in return for each family providing and equipping a soldier.21 This somewhat romantic idea of the middle Byzantine empire resting on the popular support of a free peasantry has been generally abandoned. The transition is now thought to have occurred after Heraclius’ reign, and change was probably gradual. Part of the problem is terminology. The word theme originally meant a military unit, and references to themes in sources relating to the seventh century may refer to military units, rather than to the land where they were stationed. But even if it seems that the reference is to territory, we cannot be sure that such a reference is not an anachronism, since our sources date from the ninth century when territorial themes were in place. As with the changes in civil administration already discussed, it is possible – indeed likely – that the two sets of arrangements overlapped; even though there are references to strat¯egoi and themes in the seventh century, there is still mention of provinces (eparchiai), governors, and use of such titles as magister militum well into the eighth century. While it is impossible to provide a detailed timescale, one can reasonably suggest that the themes developed in the following way. After the Byzantine army was defeated by the Arabs, the troops retreated over the Taurus mountains into Anatolia. The years following the defeat saw continual raiding by Muslim forces into Anatolia, leading finally in the 660s and 670s to a concerted attempt by Mu‘awiya to advance across Asia Minor and take Constantinople (see above, pp. 232–3 and below, p. 372). In this prolonged crisis, the Byzantine armieswere stationed in the provinces of AsiaMinor. They would have been provisioned in the traditional way, by a levy raised by the local governors from the civilian population. The areas that came to be called the themes of the Armeniakoi and the Anatolikoi were the groups of provinces where the armies commanded by the magistri militum per Armeniam and per Orientem took up their stands. The theme of the Thrakesioi covered the provinces in western Anatolia to which the army of the magister militum per Thraciam withdrew after the Arab victories in Syria and Palestine. The theme of the Opsikion was made up of the armies of the magistri militum praesentales, some of which had probably long been established in the area just across the Bosporus from Constantinople. The name derives from the title of the officer (comes Obsequii) who, during the reign of Heraclius, was appointed to command the praesental armies on the emperor’s behalf. The Karabisianoi, the fleet, formed part of the old quaestura exercitus, probably based at Samos (see also below, p. 267). It seems likely that the army corps took up the positions into which they would become embedded sometime around the middle of the seventh century. Why and at what stage the civil administration declined, to be replaced by the military government of the strat¯egoi, is much less clear. Presumably the overriding need to supply a standing army, together with the decline of the ancient, city-based economy, meant that the strat¯egoi, backed up by the increasingly powerful officials forming part of the imperial court, gradually assumed the functions of the old governing elite. The elite lost much of its raison d’ˆetre because of the court-centred nature of the civil administration.