The hydrological cycle can influence climate through a great variety of processes. A good representation of the hydrological cycle in climate models is therefore crucial. Attempts to analyse the global hydrological cycle are hampered by a deficiency of suitable observations, particularly over the oceans. Fully coupled general circulation models are potentially powerful tools in interpreting the limited observational data in the context of large-scale freshwater exchanges. We have looked at large-scale aspects of the global freshwater budget in a simulation, of over 1000 years, by the Hadley Centre coupled climate model (HadCM3). Many aspects of the global hydrological cycle are well represented, but the model hydrological cycle appears to be too strong, with overly large precipitation and evaporation components in comparison with the observational datasets we have used. We show that the ocean basin-scale meridional transports of freshwater come into near balance with the surface freshwater fluxes on a time scale of about 400 years, with the major change being a relative increase of freshwater transport from the Southern Ocean into the Atlantic Ocean. Comparison with observations, supported by sensitivity tests, suggests that the major cause of a drift to more saline condition in the model Atlantic is an overestimate of evaporation, although other freshwater budget components may also play a role. The increase in ocean freshwater transport into the Atlantic during the simulation, primarily coming from the overturning circulation component, which changes from divergent to convergent, acts to balance this freshwater budget deficit. The stability of the thermohaline circulation in HadCM3 may be affected by these freshwater transport changes and this question is examined in the context of an existing conceptual model.