A common feature to reactor containment programmes is the use of detailed models to furnish data for design and safety assessment purposes. Despite the great strides which have been made in computational methods it is expected that the experimental approach will have a continuing role. It is therefore still pertinent to review the basis of such experiments, to see how they could be improved, and to see how well model experiments describe other processes occurring during an hypothetical core disruptive accident (HCDA). Numerous papers have described experiments on detailed models of a fast reactor scheme, and in all these, the sodium coolant of the reactor is replaced by water in the model for obvious practical reasons, but the scaling consequences of this change seem to have been given little attention. Therefore the object of this paper is to review the fundamentals of the scaling process, and then to discuss in more detail the effects of changing the working fluid in HCDA experiments. It is shown that the usual practice of using a geometrically scaled model, water as the working fluid, and a charge of the same characteristics as expected in the reactor excursion results in an inexact simulation, requiring somewhat uncertain corrections before the data can be used for the reactor case. An alternative possibility which is discussed in this paper would be to model the compressible characteristics of the sodium and the results could then be applied directly to the reactor scale using well defined scaling factors. This proposal, however, does require detailed changes to the experimental model and to the charge, but neither of these is expected to give undue difficulty. Modelling of an HCDA normally refers to modelling of the compressible fluid/structure interaction but in recent years interest has grown in other processes, such as heat and mass transfer. By looking at the appropriate dimensionless numbers in the model and reactor, the possibilities of using scale experiments to investigate certain features can be gauged. It is concluded that with experiments using water as the working fluid many processes associated with heat and mass transfer will not be modelled correctly and therefore special experiments have to be devised. For the same reason, caution should be used in extrapolating to the reactor heat and mass transfer data from experiments designed to reproduce structure deformation and loading. Although the modelling of compressible fluid/structure interactions is without doubt the main interest at the present time, other processes can be modelled without difficulty. In the example given, it is shown that buoyancy effects can be modelled provided an incompressible fluid simulation is sufficient. This simulation requires a low pressure charge such as might be provided by the evaporation of FREON released from a frangible container.