This dissertation concerns two real problems of instability on the banks of major rivers. The two rivers, the Thames and the Mississippi, are both extensively canalised. Instabilities may develop in the river banks at times of flood and threaten to re-establish the natural regime of the rivers with a comcomitant risk to life and installations behind the flood defences. The two problems of instability were selected for this study both for their practical importance, and for their illustration of a theme: the interaction of two layers of soil. In each case the river bank instability may take alternative forms depending on the relative geometry of the soil layers involved. The objective of the investigation was to show that the modes of failure which are observed in the field may be understood in terms of the mechanisms by which the two layers interact with one another during failure. The instability that affects the Thames flood embankments occurs when they are subjected to flood water levels combined with high uplift pressures in a permeable layer which underlies the marsh upon which they are built. The most extreme loading occurs when a storm flood surge coincides with high spring tides . The investigation reported in Part I of this dissertation was designed to discover the nature of the failures which result from different extreme combinations of self weight and uplift loading on such an embankment. The experimental technique used throughout the investigation was that of centrifugal modelling using the large new Geotechnical Centrifuge at Cambridge. A series of 1/50 scale model embankments was constructed and tested to failure on the centrifuge. Each model was brought into equilibrium a t 50 gravities and then subjected to a planned sequence of rapid perturbations of loading. Both self weight of the embankment and uplift pressure in the foundation were varied. The load combinations were determined which caused failure of the embankments. The geometry of the embankment was kept constant in all tests, but the depth of the underlying marsh was varied. By selecting suitable parameters the failure load combinations are shown to plot on an interaction diagram, the shape of which depends on the relative depth of the embankment and the marsh upon which it is built. The shape of the failure surface which developed was observed to depend on the nature of the loading which causes the collapse. The experimental investigation is supported by a simple analytical derivation of the shape of the interaction diagram. The analysis uses material properties relevant to the centrifuge models but is, in principle, applicable to the prototype embankments along the Thames. A preliminary analysis is also presented that accounts for the dependance of the form of failure surface on the nature of the loading. The instability that affects the Mississippi river banks is caused by intense erosion of the channel bed at times of high river discharge. ~fuere the bank is composed of fine grained 'point bar' deposits of a certain type, follows tides sometimes develop which rapidly destroy large portions of the river bank, as shown in the frontispiece'. The normal explanation for the ability of a sand layer to generate a flowslide has been that the sand · which is assumed to be initially loose, liquefies and flow way. It is shmY in this dissertation that a 'flowslide' scar may develop in a sand layer which is initially dense and that the feature that: governs whether a particular failure will or will not become a flowslide, is the ability of the sand to flow unimpeded from the scar, and not its ability to liquefy. An analysis is developed which, with several unresearched assumptions, predicts the final geometry of a failure by this mechanism. On the Mississippi the agency which is able, in certain circumstances, to impede the flow of sand, is the cohesive overburden. The interaction of the two layers was researched experimentally using a series of centrifuged models. The influence of geometry on the mechanisms 'of deformation was' investigated by varying the relative thickness of the two layers. It was found that two alternative mechanisms of deformation of the clay overburden can result from the creation of an erosion crater in the sand layer. These distinct mechanisms correspond to the two types of failure observed along the Mississippi: shear type and flow type failure, and they are governed simply by the relative thickness of the overburden and sand layers. This geometrical relationship is justified analytically in terms of the ability of the overburden to arch across the crater.