This paper discusses two recent developments in the formal study of argumentation-based inference: work on preference-based abstract argumentation and on classical (deductive) argumentation. It is first argued that general models of the use of preferences in argumentation cannot leave the structure of arguments and the nature of attack and defeat unspecified. Then it is claimed that classical argumentation cannot model some common forms of defeasible reasoning in a natural way. In both cases it will be argued that the recently proposed
framework for structured argumentation does not suffer from these limitations. In the final part of the paper the work of Marek Sergot on argumentation-based inference will be discussed in light of the preceding discussion.