The thesis evaluates the role of the central bank in preserving financial stability and analyses the consequences for the structure of banking supervision and payment systems. The first chapter examines whether there is a need for a lender of last resort to maintain systemic stability. The literature on the concept of lender of last resort is critically assessed. The crucial issue is whether there is contagion risk in banking. A model is constructed to test for contagion risk. The results indicate that there is contagion risk in banking. An initial failure could generate further failures without intervention by the authorities. There is, therefore, still a role for the central bank as lender of last resort to assist ailing banks, whose failure may have a systemic impact. The second chapter investigates whether it then follows that the central bank should conduct banking supervision. The main argument for separating the functions of monetary policy and banking supervision is that combining them might lead to a conflict of interest. An argument against is that separation is inconsistent with the central bank's concern for systemic stability. In a cross-country survey of 104 bank failures, a trend towards using tax-payers' money for bank rescues is observed. This strengthens the case for assigning the supervisory function to a government agency. But it would be difficult to have a complete division, since the central bank generally remains the only source of immediate funding. The final two chapters deal with interbank payment systems. The third chapter reviews existing payment system arrangements and highlights their shortcomings. Payment systems are the key channels for the spread of systemic risk. It is found that central banks are currently the implicit guarantors of payment systems. The fourth chapter presents a model to analyse the alternatives for reducing the fragility in payment systems. The first is private loss-sharing in case one (or more) of the participants fails to settle. A methodology to measure the cost of loss-sharing is developed. Alternatively, banks can move to gross settlement, but banks need collateral before making payments. The trade-off between collateral holdings and payment delays is incorporated in the model. The results indicate that the estimated cost of gross settlement exceeds the expected value of settlement and systemic risk in net settlement. 2