Previous work with columnar castings of lead has shown that there exists a connexion between the solidification texture and the morphologies of the decanted freezing interfaces. This has been explained in terms of crystal growth from the melt by the extension of relatively large (one micron) steps or platelets which are visible on many decanted interfaces. The present work consists of an examination of preferred orientations and interface morphologies in castings of lead and of metals having structures other than face-centred cubic. Metal castings have been prepared in which the orientations of columnar grains could be correlated directly with the morphologies of the corresponding decanted solid/liquid interfaces. Experiments have been carried out with lead, zinc, magnesium, tin, bismuth and a $\beta$-brass alloy, and except in the last case, the morphologies of the freezing interfaces were controlled by varying the purities of the metals for given conditions of casting. Except in the case of tin the preferred orientations are determined by the morphologies of the freezing interfaces. The results can be satisfactorily explained if the mechanism of crystal growth is one of edgewise extension of closely packed planes in directions lying close to the plane of the actual solid/liquid interface. The origin of the platelet structure on decanted interfaces is discussed, and the appearances of decanted interfaces are compared with structures visible on the free surfaces of rapidly frozen metal sheets. It is suggested that there may be some connection between the platelet formation on decanted interfaces and those seen on the solid/gaseous interfaces, and that the existence of a residual liquid layer over the freshly decanted surfaces might give rise to a structure which is not typical of the original solid/liquid interface. The thickness of this residual liquid film is $\sim$ 20 $\mu m$, but under the conditions of these experiments the liquid freezes in less time than is required for the nucleation of new grain structures or even of eutectic lamellae. It is concluded that decanted surfaces are probably an approximate representation of the true solid/liquid interfaces, but that it is uncertain how far this is true of the platelet structure.