Modern liberal democracies pride themselves on protecting certain fundamental rights for individuals. In the context of the criminal law, perhaps the most prominent is the right of accused persons to be presumed innocent until proven guilty of criminal wrongdoing. However, similarly prominent is the phenomenon of imposing a burden of proof on the accused, the effect of which is to require them to prove their innocence. This thesis explores the relationship between these two features of the criminal law and examines how it is possible to reconcile the presumption of innocence (PoI) with ‘reverse’ burdens of proof. In particular, the thesis sets out to answer the following question: under what circumstances, if any, will it be coherent with the PoI to impose a reverse burden of proof on the accused in a criminal trial? In response, the thesis ultimately argues that reverse burdens and the PoI cannot be reconciled where the reverse burden amounts to a requirement to prove innocence. This is because, properly understood, the PoI requires proof guilt beyond a reasonable doubt. This, in turn, is understood in terms of the allocation of risk of failing to persuade the factfinder of a certain fact. Specifically, the thesis contends that the PoI prohibits the allocation of this ‘risk of non-persuasion’ to the accused, meaning that the accused must not be required to prove that they are innocent. This is a unique interpretation of the PoI in the context of reverse burdens. The thesis therefore concludes that the only reverse burdens which can be reconciled with the PoI are those which do not require the accused to prove their innocence, a set of defences known as ‘non-exculpatory defences’. Otherwise, the PoI prohibits reverse burdens. In reaching this conclusion, the thesis proceeds in three main parts. The first expands on and contextualises the problem which reverse burdens pose in terms of PoI compatibility. The chapters comprising this first part of the thesis offer a critical account of the prevailing approaches to analysing reverse burdens in several common law jurisdictions. The thesis then identifies a distinction which has often gone unacknowledged in these conventional accounts of reverse burden compatibility. This is the distinction between the PoI as a trial rule (the ‘thin’ PoI) compared to the PoI as a general norm of the criminal law (the ‘thick’ PoI). Departing from the tradition of choosing either the thick or thin PoI as the only correct version of the PoI, the thesis argues that both must be relevant to a holistic assessment of how to allocate the burden of proof in a way which is coherent with the PoI. The second part of the thesis fleshes this idea out and develops an account of the thick PoI as entailing a prohibition on requiring the accused to prove any fact which bears on their responsibility for the offence in question. In these chapters, the thesis explains why this understanding should be preferred to three alternative approaches to connecting the PoI and reverse burdens. These alternatives are: i) ignoring the thick PoI and characterising the PoI as nothing more than a trial rule; ii) assessing each reverse burden on an ad hoc basis, by reference to a requirement of proportionality; and iii) flatly banning any persuasive burden on the accused. Ultimately, it is argued that each of these is flawed for a different reason. Finally, the third part of the thesis examines the ramifications of adopting this new understanding of reverse burdens for the structure of the modern criminal law in systems such as England and Wales or the United States. The chapters forming this third part offer a new way forward, whereby burdens on the accused are generally not full persuasive burdens, and where criminal offences are drafted more carefully with the burden of proof in mind. This charts a path for a criminal law which more fully respects the rights of individuals.