The shape of the capsid of phage G, active against B. megatherium, has been studied by electron microscopy using a variety of approaches including negative staining of whole particles, examination of stereoscopic micrographs, shadowing, observations on capsids devoid of their DNA content ('ghosts'), short and prolonged fixation and the use of single and serial ultra-thin sections. The results were conflicting. With negative staining the appearance of the capsid was not consistent with an icosahedral shape but could be related to that of an octahedron. The shape of the capsid in 'ghost' preparations was definitely octahedral. However, after fixing for 1h or longer and shadowing, the capsids were clearly icosahedral in shape with their sides equal to about half the length of those of the octahedron seen by negative staining. The same results were obtained by the thin sectioning technique. From a study of the two types of geometrical figures it was found that an octahedral container can be transformed into an icosahedral one with sides of half the original length if it is folded inwards on itself at 6 consecutive points. If this type of effect occurred during the preparative procedure it would account for the two apparently conflicting appearances of the capsid. Such an effect would involve a marked change or 'jump' through an intermediate unstable form. While it would reconcile the different results it is emphasized that such an interpretation is essentially speculative. To illustrate the complexities of the problem, an interesting geometrical exercise is presented in the Appendix.