This thesis challenges one of the fundamental assumptions about Rome’s political upheaval in the third century CE. This period is conventionally defined by the growing political influence of the army at the expense of the Senate, after the Severan emperors made it clear that their hold on power rested on military support. The soldiers would grow bolder in asserting their position, eventually coming to overthrow emperors at will. I present a broad reassessment of the evidence for a historical model which derives from the narratives of two contemporary witnesses, Cassius Dio and Herodian.

Dio is the subject of my first discussion. I address two problems. Firstly, Dio’s contemporary history survives only through Byzantine epitomes and excerpts. Its irreparable alteration means that Dio’s later books cannot be treated in their own terms, but have to be contextualised against the wider thematic framework of his thousand-year account. Secondly, I turn to Dio himself. Within that framework, Dio presents himself as the ideal senatorial historian. In doing so, he is able to define a uniform senatorial experience, which excludes everything else as deriving from military corruption.

An analysis of Herodian follows, also in two parts. The first analyses Herodian’s construction of Roman society into three constituent parts, Senate, army and people. I show how these simplistically homogenous social units allow Herodian to explore imperial character, even as they cause inconsistencies in his political narrative. I then address Herodian’s account of Maximinus Thrax. This narrative has been presented as the historical culmination of the army taking over politically. I argue instead that it represents the climax of Herodian’s rhetorical scheme.

Overall, the model of political conflict is built on two contemporary accounts which have specific reasons to simplify matters in their presentation of political activity. In order to understand the nature of political change in this period, I argue that it is necessary to move beyond them.