The UK’s bridge stock is deteriorating due to a number of mechanisms that affect concrete durability. These include carbonation, thaumasite sulphate reaction, alkali-silica reaction, chloride ingress, freeze-thaw action and fatigue. There are limited resources available to maintain the bridge stock in a safe and operable condition. It is therefore essential that available funds are used to optimum effect for inspections, maintenance, assessment, repair or replacement. If a short term view is taken to bridge management then it is highly likely that a future backlog of essential work will build up which will result in severe financial and operational difficulties. In this context, methodologies that can predict the time-dependent deterioration of RC bridges would be highly beneficial to bridge owners and maintaining agents when developing bridge management strategies. A methodology has been developed in this thesis to estimate the time-dependent deterioration and structural reliability of RC components. Only chloride induced deterioration was modelled because it was found to be the dominant mechanism in the UK. Probabilistic techniques and spatial analysis are used to model the uncertainty inherent to chloride induced deterioration. Code based limit state equations are used with the deterioration models to create time-dependent failure margins. Standard reliability methods (FORM/SORM) are then applied to estimate failure probability time profiles over the life of the bridge. Furthermore, classical and Bayesian statistical techniques are used to update deterioration/reliability predictions. The deterioration models produce plots of area initiated or delaminated as a function of time. These may be used to plan inspection times and types. The ensuing limit state reliability analysis produce probability of failure profiles which may be used as a comparative tool for assessing similar bridge/component types, and thus a bridge management strategy can be developed on the basis of perceived criticality. The updating techniques are used to illustrate how model predictions can be improved when site data becomes available. It is shown that the development of models using typically available site data can be beneficial to bridge owners, but the quantity of such data needs to be enhanced. This is particularly evident for the wide diversity of bridge structures in the UK.