John Burnheim, in Is Democracy Possible? (1985), rejects the present system of electoral democracy and the idea of direct participation by all citizens in every aspect of public decision-making. Instead, he advocates a decentralised form of democracy which incorporates broader citizen participation than that of contemporary liberaldemocracies. This alternative proposal, which Burnheim calls "demarchy," makes use of the ancient Athenian principle of choosing by lot the representatives who are to hold various public offices. The determining features of demarchy are: decentralisation of government consisting of specialised functional authorities, each independent of each other, with specific powers over specialised problem areas; coordination of the relevant decision-making bodies through mutual negotiation and accommodation; and statistical representation of the different interest groups affected by the decisions.In this thesis I contend that although demarchy eliminates a large number of the problems peculiar to electoral democracies many questions still need to be addressed. In particular, I argue that the sampling procedure Burnheim proposes, namely, statistical representation, needs to be re-evaluated. My reasons are twofold. Firstly, as a solution, statistical representation does not follow from Burnheims own criticisms of representative democracy and his arguments for functional democracy. Secondly, it fails to provide a flexible solution to the problem of representation. I offer a modified version of Burnheims original proposal for the democratic reorganisation of society, which also follows the principles of rotation of offices, deliberation and participation by lot. However, my proposal, which I refer to as "cross-section demarchy," is based on unconditional random selection and not on a statistical sampling procedure. Unconditional random selection by lot, unlike statistical representation or elections, offers a solution for measuring popular participation in social and political decision-making. Moreover, I maintain that participation by lot on a rotating basis redefines democracy in terms of "ruling and being ruled in turn," and shifts the emphasis away from representation or direct participation by all in everything.I also claim that a viable theory of democracy must incorporate a theory of education which is consistent with the general principles of that democratic proposal. I recommend the implementation of an established model of education by Matthew Lipman, which promotes the preparation of citizens for participation in a democracy. Drawing attention to a recent debate between the value of "education for democracy" and "democratic education," I contend that if the role of education is to produce an educated citizenry capable of making political judgments, then democratic education is essential for the preparation of such a citizenry.