In previous work, a modelling methodology was developed to determine statistical information, similar to that which can be obtained from H-1 MRI, on the spatial distribution of different pore sizes within porous media using mercury porosimetry. The new methodology has the advantage over MRI that it is suitable for application to chemically heterogeneous materials of interest in catalysis, such as coked catalysts and supported metals, that are not amenable to quantitative studies using conventional 1 H MRI. However, the new methodology relied upon the theory that the entrapment of mercury within many porous solids, such as sol-gel silicas, occurs because the spatial distribution of pore sizes within the material is not random, but is, in fact, highly correlated. Mercury entrapment is thought to occur in isolated, macroscopic (>10 mum in size) domains, containing similarly sized larger pores, that are completely surrounded by continuous networks of smaller pores. In this work mercury porosimetry experiments on silicas, consisting of a primary intrusion and retraction cycle, followed by re-injection of mercury and secondary retraction, have shown that the entrapment of mercury does not arise due to the alternative possibility of irreversible pore structural collapse. Light microscopy studies of transparent samples following mercury porosimetry experiments have confirmed that the spatial distribution of entrapped mercury is highly heterogeneous and occurs in particular macroscopic regions within the sample. These findings support the underlying theory of mercury entrapment used in previous simulations of the mercury porosimetry experiment. (C) 2003 Elsevier Science B.V. All rights reserved.